~1
Water Demand Management in Kuwait
by
Milan Milutinovic
B.S. Civil Engineering
University of Belgrade in Serbia, 2004
SUBMITTED TO THE DEPARTMENT OF CIVIL AND ENVIRONMENTAL
ENGINEERING IN PARTIAL FULFILLMENT FOR THE DEGREE OF
MASTER OF ENGINEERING IN CIVIL AND ENVIRONMENTAL ENGINEERING
AT THE
MASSACHUSETTS INSTITUTE OF TECHNOLOGY
INSTTE
OF TECHNOLOGY
MASACHUSETT
JUNE 2006
JUN 0 7 2006
( 2006 Massachusetts Institute of Technology
All rights reserved
LIBRARIES
BARKER
Signature of Author..........................
)epartment of Civil and Environmental Engineering
May 12, 2006
C ertified by....................................
..
........... ............
. . . . . .. .
Elfatih A. B. Eltahir
Professor of Civil and Environmental Engineering
Thesis Supervisor
qfT t I
A
Accepted by.....................................................................Whittle
Chairman, Departmental Committee for Graduate Students
Water Demand Management in Kuwait
by
Milan Milutinovic
Submitted to the Department of Civil and Environmental Engineering
on May 19, 2006 in Partial Fulfillment of the
Requirements for the Degree of Master of Engineering in
Civil and Environmental Engineering
ABSTRACT
Kuwait is an arid country located in the Middle East, with limited access to water resources. Yet
water demand per capita is much higher than in other countries in the world, estimated to be
around 450 L/capita/day. There are several reasons for such a high demand, but one is certainly
the price. Water does have its pricing schedule in Kuwait, but in reality water bills are not
collected. The main objective of this thesis is to investigate the impact of water pricing as a tool
for managing water demand. The original idea, to construct a water demand model for Kuwait,
was modified because of the lack of data about the effect of price increases and household water
consumption characteristics in Kuwait. So, water demand models described in the literature for
several arid regions were adapted and recalibrated for Kuwait. Simulations describing the
influence of block tariffs, constant prices, free allowances followed by various pricing schemes
were conducted. A pricing schedule has been proposed that consists of a free allowance followed
by a constant price. The proposal has the following logic: if water is consumed wisely, only to
satisfy vital needs, it should be free. However, to limit over consumption, the quantity of water
over the allowance should be priced. The results showed that this kind of pricing schedule would
be efficient in significantly reducing demand. The models show that a price of water of $1/M 3,
after a 150L/capita/day allowance, would reduce the demand by about 35 percent (with a range
of around 20-40 percent, depending on the demand model used).
Thesis Supervisor: Elfatih A. B. Eltahir
Title: Professor of Civil and Environmental Engineering
Acknowledgments
Special thanks to my thesis advisor, Professor Eltahir, for his outstanding guidance
and expert advising. I would also want to thank Professors Rafael Bras and Kenneth Oye, and
all others who have worked on the Kuwait projects, for the excellent suggestions they gave.
Furthermore I would like to thank my program advisor Dr. Eric Adams for all the
excellent advices and support during the year. Also, thanks to my colleagues and friends at
MIT for their support. I would like to express special thanks to my family and friends back in
Belgrade.
Table of Contents:
L ist of T ables..............................................................................5
L ist of Figures..............................................................................6
1. Introduction ..............................................................................
7
2. Water resources in Kuwait.........................................................9
3. Literature review
3.1 Water Demand Models..................................................14
3.2. Water metering.........................................................29
3.3. Water U sage................................................................31
4. Price
Proposal................................................................................
33
5. Simulations of the Influences of Various Other Pricing Schemes.............66
6. C onclusion.........................................................................
7. R eferences..........................................................................85
4
84
List of Tables
Total capacity in Kuw ait.....................................................................11
Total capacities for various desalination processes....................................12
Multistage Flash Plant Capacities.......................................................12
Variables and coefficients used in the model by Zhou and Tol (2005)
for calculating desalination costs...........................................................13
Table 3.1: Income and household size elasticity from various studies..........................17
Table 3.2: Summary of price elasticities in some studies of residential water demand..... 21
Table 3.3: Influence of non-price policies..........................................................23
Table 3.4: Elasticities of low-flow showerheads...................................................24
Table 3.5: Elasticities computed using the Stone-Geary Form...................................27
Table 4.1: Price elasticities from studies of residential water demand..............................40
Table 4.2: GDP per capita in USA and for Kuwaiti and Non-Kuwaiti..........................41
Table 4.3: Household income USA...................................................................42
Table 4.4: Assumed Household income for Kuwaiti and non-Kuwaiti..........................42
43
Table 4.5: Households in Kuwait ....................................................................
Table 4.6: Household size distribution used in the simulations......................................43
Table 4.7: Coefficients used in the Saudi Arabia model...........................................44
Table 4.8: Values of some of the coefficients in the California Model..........................47
50
Table 4.9: Coefficients used in the Static Annual Australian Model ............................
Table 4.10: Income coefficient values for adopting the Australia Model.......................51
Table 4.11: Water demand equation coefficients for the Tunis model...........................54
Table 4.12: Consumers proportion equation for the Tunis model...............................54
Table 4.13: Changed coefficients for adopting the Tunis model.................................55
Table 4.14: Coefficients used in the Stone-Geary model.........................................56
Table 4.15: Elasticities obtained in the simulations and in the original models................62
Table 4.16: Reduction in demand in simulations at the marginal price of $1/i 3 ..... . . . . . . . . . . 64
Table 5.1: Original coefficients and coefficients used in the California Model
71
block tariff simulations....................................................................
model.............................................72
Table 5.2: Block tariffs used in the California
Table 5.3: Coefficients used in the Tunis model water demand equation for
77
block tariffs simulations....................................................................
Table 5.4: Coefficients used in the Tunis model consumers proportion equation for
block tariffs simulations..................................................................77
Table 5.5: Block tariffs used in the Tunis Model Simulations......................................77
Table 5.6: Coefficients used in the California Model simulations of the influence
of the percentage of the household bill on demand......................................83
Table 2.1:
Table 2.2:
Table 2.3:
Table 2.4:
5
List of Figures
3.1: Non-agricultural use of water.............................................................31
3.2: Dom estic w ater use............................................................................32
4.1: Price Proposal - A free allowance followed by a constant price.......................36
4.2: Locations here the models were estimated.............................................43
4.3: Decreases in demand for selected models for
a constant price schedule without an allowance......................................58
Figure 4.4: Monthly government subsidy in Kuwait for a constant price schedule
59
w ithout an allow ance.....................................................................
Figure 4.5: Decreases in demand for selected models for an allowance
followed by a constant price schedule....................................................61
Figure 4.6: Monthly government subsidy in Kuwait for an allowance
followed by a constant price schedule................................................61
Figure 4.7: Influence of the allowance and price on demand reduction
for the Australian Model.................................................................65
Figure 4.8: Influence of the allowance and price on demand reduction
for the Saudi Arabia Model..............................................................65
Figure 4.9: Computed distribution of water demand for a free water situation.................66
Figure
Figure
Figure
Figure
Figure
3
Figure 4.10: Computed distribution of water demand for a price of $1/M .......
. .. . .. . . . . . . . .
67
Figure 4.11: Computed distribution of water demand for a price of $2/M ....... . .. . . . . . . . . . . . 67
Figure 5.1: Block pricing schedules used in the California simulations...........................72
Figure 5.2: Influence of block tariffs for the California model...................................74
Figure 5.3: Influence of pricing on Kuwaiti (California model).................................74
Figure 5.4: Influence of pricing on Non- Kuwaiti (California model)..............................75
Figure 5.5: Influence of constant pricing and a block tariff system
on the Total Cost, Price and Government Subsidy for the California model........75
Figure 5.6: California model- influences of variables in the demand equation...............76
Figure 5.7: Block tariffs used in the Tunis Simulations.............................................78
Figure 5.8: Influences of block tariffs on demand in the Tunis simulations.....................79
Figure 5.9: Responds of the Higher bracket in the Tunis Model.................................80
Figure 5.10: Responds of the Lower bracket in the Tunis Model...............................80
Figure 5.11: The total cost and price of water and government Subsidy
81
in the Tunis model......................................................................
model..............................................82
Figure 5.12: Shifting between block in the Tunis
Figure 5.13: Block Tariffs constructed to have a certain percentage of the total income........84
Figure 5.14: The percentage of income paid for water...........................................85
Figure 5.15: Government subsidy for various income percentages.............................85
3
6
1. Introduction
The quantity of water consumed in Kuwait is estimated to be around 450 L/capita/day
(Darwish et al. 2005). When compared to other countries (USA - 333 L /capita/day (EPA);
France 164 L/capita /day; Germany 127 L/capita/day (EWA 2005 yearbook)), a need to
decrease demand is evident. There are several reasons for such a large demand (temperature,
pipeline leakage), but certainly one reason is the fact that water bills are not collected
efficiently. Different aspects of water demand management have been studied and an overall
approach has been proposed to address Kuwait's large water demand problem.
First, a review of the water resources in Kuwait is presented. Particular attention is
dedicated to desalination, since Kuwait gets over 90% of its potable water through desalting
plants (Hamoda 2001). The cost to produce water in Kuwait is high, estimated to be around
$3/M
3
(Darwish and Al-Najem, 2005)
Next, a literature review of areas that are related to water demand was conducted. A
broad review of water demand models found in the literature is presented. Most of the
demand models suggest that water consumption is elastic to price increase, with a wide range
for elasticity, typically between -0.1 and -0.8. Also a report about water metering and water
usage is included. It has been found that water metering has two main advantages: the
conservation of water is rewarded, and metering has shown to reduce demand (Lund 1988).
Also, studies have shown that around 30% of domestic water use goes for toilets, and about
20% for showering.
The fourth and fifth chapters consist of simulations of water demand in Kuwait using
various pricing schemes.
The original idea was to develop a water demand model for
7
Kuwait. However, with the lack of data regarding household water characteristics (water
consumption is generally not metered in Kuwait) and the influence that a price increase has
on demand (there has not been an increase in the price of water in recent years), the
perspective of the research was modified, and water demand models from the literature were
adopted for Kuwait.
Chapter four presents the main result of the research. It consists of a price proposal, a
free allowance followed by a constant price. The main objective of this proposal is to
eliminate the waste of water by pricing it after a certain amount, which would satisfy the
basic needs. The quantity of the allowance was proposed to be 150 L/capita/day, an average
consumption found in European countries. Simulations of the water demand showed that this
pricing schedule would eliminate the waste of water and reduce consumption by about 35%
for a price of $1/M
3.
Chapter five presents the simulations completed prior the proposal. They
were conducted to study the demand models and the influence that various pricing schemes
had on demand.
8
2. Water resources in Kuwait
Kuwait is an arid country located in the Middle East, rich in oil but poor in water
resources. The annual average rainfall is around 110 mm, with a variation of 31mm to
242mm. Surface runoff and groundwater recharge from rainfall are rare. This is primary due
to factors such as the high potential evapotranspiration rate (Fadlelmawla and Al-Otaibi,
2005).
The main water resources in Kuwait are brackish groundwater, desalination and
treated wastewater. Wastewater is treated in three plants in Kuwait to a tertiary level, with an
average flow through the plants of 388,000 m3/day (1999). Depending on the season, between
25-40% of this water is used for irrigation, while the rest is disposed into the sea (Hamoda
2001).
The natural recharge to the groundwater aquifers is smaller than the extraction rate.
Fadlelmawla and Al-Otaibi, 2005 cited that the annual groundwater production in 1999 was
around 118 million m3, while under steady-state conditions, the recharge, generated through
lateral flow from Saudi Arabia, was estimated to be 44 million m 3/year. Special interest in the
thesis was dedicated to desalination in Kuwait, since over 90% of potable water in Kuwait is
acquired from desalination (Hamoda 2001).
Desalination
Desalination Technologies
The total desalination capacity of the world is around 37.5 million m 3 /day (Abu-Arabi and
Reddy, 2005). Based on data from the American Water Works Association (2004) the most
frequent desalting technologies are Multistage Flash (44%) and Reverse osmosis (40%),
9
while Electrodialysis represents around 6% of the total capacity. Generally, the processes of
desalination can be divided into two main categories: membrane processes (i.e. Reverse
Osmosis, Electrodialisys) and thermal processes (i.e. Multistage Flash).
Reverse Osmosis is the dominant approach in the USA. During the process, dissolved
ions and water molecules are separated by the use of semipermeable membranes. Based on
the dissolved solids' concentrations, desired recovery, and specific membrane performance, a
pump provides the required pressure. The technology does not require as large energy
consumption as the thermal processes. Electrodialysis is used for salinities smaller than 2000
mg/l, and mainly for smaller facilities.
Membrane processes are mostly used in countries with low energy costs (Middle
East). In the Multistage Flash technology (MSF) water is heated in a brine heater, then flows
through a number of vessels (stages), where the pressure is lowered, causing water to
immediately boil. In every next stage, the pressure is lower, so no additional heat needs to be
added. Frequently the MSF plants are combined with electric power plants to reduce energy
consumption. Multiple Effect Distillation (MED), similar to the MSF, has a number of
vessels with variations in pressure in temperature. MED has a higher performance ratio and
works with lower temperatures than the MSF (AWWA, 2004)
Costs
The cost to desalinate water depends on the type of desalination process, energy costs,
and labor cost. The AWWA (2004) presents the next information concerning costs.
The costs can be divided into Capital and O&M costs. The Capital costs are
construction and incidentals costs. Capital costs are a function of the quality of the source of
10
water, production goals, capacity (bigger capacity- smaller cost), brine disposal, etc. The
typical construction costs for a 19,000 m 3/day capacity plant are $0.8-$1.5 per 1/day installed
cap for a sea-water desalting plant and $0.3-$0.5 per L/day installed capacity for a brackish
water plant. Operation & Maintenance cover the costs of power, labor, chemicals, membrane
replacement, etc. A sea-water desalting plant (19,000m 3 /day capacity) would have a
desalination water cost of $0.4-$0.65 per m3 , while a brackish water plant would obtain a cost
of $0.15-$0.3 per m 3 .
Abu -Arabi and Reddy (MEDRC, 2005) cite that for the Multistage flash, the capital
cost would be US$ 1.3 to 1.6 per installed L/day, while the desalinated water cost would be
US$ 0.8 to 1.5 per M 3 . For the MED the authors present that the capital costs would be $0.91.2 per installed L/day and O&M costs of $0.7-1/m3. For other technologies they show
smaller expenses. Wangnick (2002) assumes that the total costs could be split up into 40% for
interest and depreciation and 60% for the running costs. However, these references do not
provide what energy cost is assumed, that is important due to the frequent changes in the cost
of oil
Desalination Capacity in Kuwait
The first desalination plant was built in Kuwait in the 1950s. Currently, Kuwait possesses 6%
of the world capacity. The following tables represent the total desalination capacities (Table
2.1), the capacities of various desalting processes (Table 2.2), and the capacities of the MultiStage Flash plants (Table 2.3), as the dominate technology.
Table 2.1:
Annual
Production rate
[M m3/year]
375.17
Total Capacity in
Average
production rate
[M m3/day]
1.023
Kuwait (Darwish et al, 2000)
Installed
Maximum
production rate capacity
[M m3/day]
[M m3/day]
1.6
1.26
11
Table 2.2: Total capacities for various desalination processes (Dr. Al-Sabeeh, 2001)
Capacity [m3/day]
Type of Plant
1,468,750
Multi Stage flash
11,672
Multi effect 150
Vapor compression
166,472
Reverse Osmosis
5,093
Electrodialysis
1,800,000
Total
Table 2.3: Multi Stage Flash Plant Capacities (Hamoda, 2001)
Plant
Shuwaikh
Shuaiba south
Doha east
Doha west
Az-Zour South
Total
CAPACITY [M m3/day]
0.082
0.136
0.437
0.192
0.328
CAPACITY [M m3/year]
29.9
49.6
159.5
70.1
119.7
1.175
428.9
YEAR
1960
1971
1978
1983
1988
1
Darwish, et al. (2002, 2004, 2005) published a number of articles about desalination
in Kuwait. He cited that all Kuwait MSF units (except three in the Shuwaikh plant) are
combined with electrical power plants to conserve energy (Cogeneration Power Desalting
Plants, CPDP). Around 10% of the oil production in Kuwait was used for power and water
production, of that 20% of the fuel that is used in the cogeneration power desalting plants in
Kuwait in 2003 was used for desalination. Darwish estimated that if the power cost is
$0.06/kWh, then the energy cost to produce one cubic meter of water is $1.32 (based on
22kWh/m 3 and a price of oil around $40/barrel). He assumed that if the energy cost is 45% of
the total water cost, the cost will be in the range of $3/M 3 (2002).
The total cost of fuel for water desalination was 612.3 million US dollars in 1999 (Hamoda,
2001)
12
Zhou and Tol (2005) show that currently water can be desalinated using the MSF
technology for 1$/M 3 . They also developed a regression model for estimating the cost to
desalinate water using various technologies. For the MSF plants, two forms of the model are
presented, a semi-log and a double log form. They present that the average unit cost to desalt
1m
3
of water is a function of the total cumulative installed capacity (TIC), the capacity of a
single plant (CAP), the contract year of the plant (YEAR), and regional (MENA) and raw
water quality dummies (SEA). TIC and YEAR are correlated, so they are not used the in the
same model (either one or the other). The regression function (2.1) has the next form when
the YEAR dummy is used for the log-log form of the model:
Log(UNITC) = ao + a, -log(CAP) + a 2 -log(Year) + a 2 -log(CAP) + MENA + SEA (2.1)
Table 2.4. - Variables and coefficients used in Zhou and Tol (2005) model
Coefficients for the
Log-Log model 2
coefficient
variable description
UNITC
average unit cost of desalting water
00
Constant
798.76
OC
CAP (Capacity of a single plant)
-0.14
OC2
YEAR (contract year of the plant)
-105.02
MENA
ME&NA (regional dummy)
0.05
SEA
SEA( Raw water quality dummy)
0.69
With input data from Table 2.3 (Hamoda, 2001), a cost to desalt water of $2/M 3 was
calculated for Kuwait Multistage Flash Plants. This big value was obtained because of the old
age of some of the plants. The model does not include changes of oil prices.
13
3.1 Water Demand Models - Literature Review
3.11. Introduction
With the increase in worldwide water demand over the last few decades, water
utilities face problems of supplying the quantity of demanded water. Water pricing, together
with other options, showed to be an efficient tool in controlling water consumption. Many
studies have researched the influence of pricing. The Journals "Land Economics" and "Water
Resources Research" have dedicated much space to this study.
A number of the studies were influenced by or used previous research developed in
the study of electricity demand (i.e. Taylor 1975, Nordin 1976). Most of the studies are
regression models based on data collected during various surveys, in regions where water
prices increased.
In a large number of water demand studies, there are many different approaches.
There is no consensus on the correct method to predict the demand for water. This is in part
influenced by the fact that every region has its own characteristics regarding water use and
socio-economic influences. Most studies find that household characteristics, water prices,
climate and seasonal changes and conservation campaigns influence price elasticity.
Water demand studies started in the 1960 and 70s mainly in the USA. In the 1980s,
the number of studies increased significantly, mostly encountering regression models based
on various data sets in water scarce areas of the US. In the 1990s, conservation methods and
water efficient technologies received more attention. Also, a number of studies were done in
European and other countries. In addition, some new methods were investigated in order to
predict the water demand.
14
This literature review presents specifications of the models, variables used, technology
changes, non-price policies, and some new studies in this field that differ from earlier
research.
3.12. Models Specification
Form
Most of the demand models are regression models. They typically use the form Q = f
(P, Z) where P are the price variables and Z are factors such as income, household
characteristics, weather, etc (Arbues et al. 2003). The most common forms are linear and
logarithmic. There is no agreement about which functional form gives better results. Some
researchers specify the form by seeing which model better fits their data set. Billing and
Agthe (1980) cite that the elasticity in the log model is more useful if the demand is a
rectangular parabola, while the elasticity in the linear form is more useful if water demand is
linear over a relevant range.
The main flaw that researchers attribute to the linear model is that at some price, the
demand for water will be zero, which is not logical as a minimum level of water consumption
is needed to survive (Arbues et al, 2003).
Estimation methods
Different estimation methods are used in the studies. The most common are Ordinary
Least Squares (OLS), Two and Three -Stage Least Squares (2SLS, 3SLS), and Maximum
Likelihood. The choice of the method is somewhat influenced by the data set that the
researcher possesses
15
Data sets
A number of different datasets have been used, ranging from individual household
data to aggregate data. A number of the studies used surveys conducted on a sample of
households (Rizaiza 1991, Dandy et al. 1997, Renwick and Archibald 1998), other
researchers used surveys conducted by the American Waterworks Association (Nieswiadomy
- 1984 survey, Foster and Beattie - 1960 survey).
Researches used cross-sectional data (Foster and Beattie 1979, Chicione and
Ramammurthy 1986, etc.), times-series data (Billings 1982), and most commonly crosssectional-times series data (Nieswiadomy and Molina 1989, Renwick and Archibald 1998,
Chicione and Ramamurthy 1986, etc.).
3.13. Variables
Household characteristics
Household characteristics are an important factor influencing water demand. All
studies include monthly household income as a significant variable that increases water
demand. In the deficiency of income data, some demand models use property value as an
alternative (Dandy et al 1997).
A number of researchers include lot size as a significant variable (Renwick et al 1998;
Dandy et al 1997; Lyman 1992, etc.). Houses with larger lot sizes are expected to have larger
outdoor water use (Renwick and Green 1998). Also, household size was frequently used in
the demand equation (Nieswiadomy 1992, Renwick et al 1998, Dandy 1997 etc.) as having
significant influence on demand. Density of households (Foster, Renwick 1998; Nauges
16
2000), the' number of faucets and age distribution of household members (Lyman 1992) are
used in some studies too. Table 3.1 presents income and household size elasticities found in
some water demand studies.
Table 3.1.: Income and household size elasticity from various studies
Author
Study area
model
Income elasticity
Hewitt and Hanemann
Texas
D/C
0.15
Renwick, Archibald 1998
California
linear
0.36
Dandy et al 1997*
Australia
Linear
Griffin et al(1990)
Texas
linear
SR: 0.14 1LR: 0.32-0.38
Household size
SR 0.04; LR 0.19
0.3-0.48
*Dandy et al. in his annual model used property value as an indicator of income
(SR -short range; LR - Long range)
Some models include lagged consumption in their models (Dandy et al. 1997,
Nieswiadomy and Molina 1991). The Dynamic model, with an included lagged consumption,
is used because water use tends to respond slowly to changes in price and other variables,
because water-using durables, like washing machines, swimming pools, etc. tend to change
only steadily (Dandy et al. 1997).
Price Variables
The most common question in the water demand literature is whether the average
price or the marginal price combined with the difference variable should be used as the price
variable in the demand equation. Although it has been the subject of a thorough debate in the
literature, a consensus has not been reached yet.
The Debate: Howe and Linawever (1967) cited that using the marginal price alone will have
invalid results in the presence of block tariffs. Taylor (1975) suggested an alternative method
17
by including two price-related variables in the estimating model, when block rates are
applied. Nordin (1976) modified it, citing that the second price variable should be the
difference between the consumers actual bill and what would be paid if all units were
purchased at the marginal price (in the case of a declining block tariffs for electricity).
Billings and Agthe (1980) implement the difference variable under increasing block tariffs
for water demand, showing that it is correct and statistically significant. Economic theory
suggests that the coefficients in front of the difference variable and income variables should
be the same magnitude, but with opposite signs. However, empirical evidence shows that the
coefficient on income and difference should have different signs, but with a bigger coefficient
in front of the income variable.
Billings and Agthe (1980, 1982) argue that the use of the average price will generate
bigger elasticities when a block pricing schedule is implied, especially when the marginal
price increases, while the intra-marginal rates remain the same. In this case the change in
marginal price is greater than the change in average price. A possible situation is that with an
increase in MP, the AP remains constant or even decreases. Billings and Agthe (1980, 1982)
also cite that the effect of a change in rates may have different effects on water use; the use of
average price alone ignores this, and produces less accurate results
In many recent studies on water demand, the MP combined with the difference
variable is used to show price elasticities (Renwick and Archibald (1998); Renwick, Green,
and McCorkle (1998); Dandy, Nguyen, and Davies (1997); Nieswiadomy and Molina
(1989)).
However, many earlier studies use the average price (Wong 1972, Young 1973,
Foster and Beattie 1979). In their studies Foster and Beattie (1981) recognize that the Nordin
18
specification (the use of MP and difference variable) was not significantly different that the
average price specification. They also emphasize questions regarding the knowledge that
consumers have on their MP and the way of block pricing and if their reaction is actually set
according to the average price.
Shin (1985) constructed a price perception model for electricity demand that describes
the response of consumers to MP or AP. He cited that the average consumer does not know
the actual rate schedule. Nieswiadomy (1992) gives reasons supporting the average price
variable because of the difficulty of determining the actual water usage during the month, as
water meters are difficult to read. In addition he cites the difficulty of knowing when blocks
have been switched and the fact sewer charges can confuse the consumer.
Shin (1985) defines the price perception parameter as P* = MP (AP/MP)
k,where
k is
the price perception parameter. Thus, if k = 0 the consumer responds only to the MP, if k = 1
then the consumer responds only to the average price. If 0 < k < 1 then the price perceived is
between AP and MP. Shin finds that electricity consumers react to average prices in his
study. Nieswiadomy (1992) tests the Shin model for water demand. His results indicate that
consumers react more to average prices than to marginal prices; k is approximately equal to 1
(although in his 1991 study he found that consumers react to marginal prices)
Opaluch (1982) also suggests a test concerning the measure of the price to which
consumers respond, for a two block tariff schedule. The hypothesis was conducted through a
thorough utility theoretical framework by Opaluch (1981). He suggests a demand equation:
±B1
QQ=B
=B+B
Q-
.PX.
2 +B 3 -(
-P+B2+' P2
')jB.(
).Q+ B4 2- (Y
pP - P2)' Q1)
1 ) where:
(P
wee
total purchases of the goods subject to block pricing
P, - price index for other relevant goods
19
Pi - price of Q in the first block
Qi - quantity of the good which is subject to the initial block pricing (PI)
Y - total income of the consumer
Q.
If the consumers react to
The average price is AP = P - Q1 + P2Q -(Q - Q1) = P2 + (P - P2)'91
Q
the block tariff schedule, then B 3
Q
=
0, and the demand equation reduces to Nordin's
specification. If the consumers react to the average price, B2 = B3 the equation uses average
price as a variable.
The Conclusion: A number of studies accept the idea that the preferences between different
price specifications are influenced by empirical rather then theoretical factors. Foster and
Beattie (1979, 1981) state that the price schedule to which consumers react should be a
subject for testing with available data. Basically, if the consumers think the water bill is
significant, they will put in the effort to learn about the pricing schedule and their exact
consumption and marginal price. Otherwise, where the water bill represents a small
percentage of income, the consumer will react to the average price (Nieswiadomy 1992, Shin
1985)
A review of accounted price elasticities and price variables used in various studies are
presented on Table 3.2.
20
Table 3.2. : Summary of price elasticities in some studies of residential water demand
Authors
Study area
Price variable
Price Elasticity
Howe and Linaweaver (1967)*
USA
AP
-0.23
Gibbs (1978)*
Miami, Florida
-0.51
Foster and Beattie (1980)
Exponential
USA
AP
-0.35 to -0.76
Billings (1982)
Lin/Log
Tucson, Arizona
MP & D
-0.66/-0.56
Schefter and David (1985)*
Wisconsin
-0.12
Chicoine et al. (1986)*
Illinois
-0.71
Linear
Illinois
MP (AP)
-0.6 on MP
Nieswiadomy and Molina (1989)
Linear
Denton, Texas
MP & D
-0.86
Griffin and Chang (1990)
Linear
USA
AP
-0.16 to -0.37
Riazaiza (1991)
Logarithmic
Saudi Arabia
AP
-0.4 to -0.78
Chicoine and Ramamurthy
(1986)
-0.10
Copenhagen, Denmark
Hansen (1996)*
Renwick and Archibald (1997)
Linear
California
MP & D
-0.33
Hoglund (1997)
Linear
Sweden
MP & AP
-0.20 on AP
Dandi et al. (1997)
Linear
Australia
MP & D
-0.63 to -0.77
Renwick, Green, McCorkle
Logarithmic
California
MP & D
-0.16 to -0.21
Nauges and Thomas (2000)
Linear
France
AP (&MP)
-0.22
Ayadi et al.(2003)
Logarithmic
Tunisia
AP
-0.17
(1998)
*
Data from Nauges and Thomas (2000)
Seasonal and Climate Variables
Most researchers found that seasonal changes and climates influence water
consumption. However, they used different variables. Billings et al. (1980,1982) use
evapotranspiration from Bermuda grass minus rainfall, Dandy et al. (1997)
use moisture
deficit (MD = PE-0.6R, where 0.6R = effective rainfall, MD = moisture deficit, but only for
the summer demand), Foster and Beattie (1981) use precipitation during growing season,
21
Ajadi et al (2003) used rainfall, while Nieswiadomy and Molina (1991) used weather as a
variable.
A Number of studies also use temperature in their models (Nieswiadomy, Renwick et
al., Riaza, etc.). Renwick et al. (1998) included the influence of temperature and rainfall in
their water demand model. Following Chesnutt and Mcspadden, they present two equations
for influences that temperature and climate have on demand. To include the influence of
seasonality these equations used sine and cosine Fourier series for the maximum daily air
temp (3.1) and cumulative monthly precipitation (3.2). These values are then included into
the demand equation.
ln(DTEMP) =Y +
ln(DPREC)=
+
v,
6 r-.
-sin(2-)
sin 2
+
7 -COS2
+ Y2;# cos 24t
}
+ej
+er
(3.1)
(3.2)
A number of studies found that summer demand is more elastic to price increase than
is winter demand (Lyman 1992, Dandy et al. 1997, Griffin and Chang, etc.). Dandy used
seasonal models (winter and summer) in his studies. Also studies have found that outdoor
water use is more elastic than indoor.
Nieswiadomy cites that in a log-log model temperature has a nonlinear relationship
with demand; the marginal impact of temperature goes up with increases of temperature; he
also cites that variations of temperature below 18C have no impact on water demand.
22
3.14. Effects of Non-Price Policy on Household Demand
Previous studies have shown that non-price policies reduce demand. Renwick and Green
(1998) showed that non-price Demand side Management (DSM) policy instrumnets have
influence on demand. In their demand equation they included six variables: Public
information campaigns (INFO), distribution of free retrofit kits (RETRO), low-flow toilet
rebate programs (REBATE), water rationing policies (RATION), water use restrictions
(RESTRICT), compliance affirmation policy (COMPLY).
In their study of California Water agencies they find that policies reduce water demand by the
percentage presented in table 3.3
Table 3.3. : Influence of non-price
policies
Logically,
more obligatory
policies reduce
Variable
%or reduction
INFO
8%
RETRO
9%
RATION
19%
As the authors
RESTRICT
29%
influenced by the quality of the implementation
COMPLY
Not significant
REBATE
Not significant
demand for water more than voluntary policies.
conclude, the outcome is
of these policies.
Nieswiadomy (1992), using experience
in Tucson cites that a campaign is successful in decreasing demand only for a few years. Yet,
after a few years use increases back to its previous level. He cites that only a major public
campaign accompanied with a price increase will have success in the long run. Nieswiadomy
also suggests that education programs will probably have more effect in water scarce regions,
because of the awareness of water scarcity.
23
3.15. Influence of Technological change on the Demand for Water
Influence of technology changes only recently became evident. Renwick and
Archibald (1998) found that increasing the number of low flow toilets in a household by one
would decreases household demand by 10%, while Chesnutt et al. (1992) found that it would
decrease the demand by 11%.
Regarding the efficiency of low flow showerheads, the next elasticities were perceived:
Table 3.4: Elasticities of low-flow showerheads
Renwick and Archibald
Whitcomb
Chesnutt and McSpadden
8%
6.4 -9.7 %
2%
Low flow toilets and showerheads reduce water by having more efficient technologies
and insure significant long term demand reduction with no required changes in the behavior
of consumers (Renwick and Archibald 1998). In the same study, they perceive that the
elasticity for adoptions of water efficient irrigation technologies for low and high density
households is 31 and 10 percent, respectively.
Nieswiadomy warns that even if a water efficient device is installed, the consumer
may react by using more water knowing about the conservation effect of the device, therefore
offsetting the conservation impact of the device.
Agthe and Billings studied effects that would make consumers install water efficient
technologies in individual households and apartments. They found that obligations to save
money, income, household size and summer marginal prices effected the decision.
24
3.16. Recent studies
Maximum-Likelihood Models
Recently, maximum-likelihood models were used to predict price elasticity (Hewitt
and Hanemann (1995), Pint (1999), etc.). Maximum-likehood models were previously
applied in the labor supply literature. These two models are specified in a two-stage
framework, they are based on likelihood functions that show the probability that a household
will choose a particular block, in a discrete way, combined with the probability of its
particular level of use in the chosen block, in a continuous way. Hewitt (1993) presented
three different maximum-likelihood models: the heterogeneous-preference model, the error
perception model and the two-error model. The models are structured based on the assumed
source of error in estimating household demand. These errors can be errors in data, missing
variables or errors in the household's actual consumption relative to its intended
consumption. The models directly allow both economic and non-economic influences, they
cite that variation in behavior is due to both price and income and influences represented by
various socio-demographic variables (Pint 1999).
However, Hewitt and Hanemann using the two-error model got higher elasticities than
in previous studies (-1.6), while Pint pointed out that elasticity is bigger in the two-error
model (-0.2 to -1.24) than in the heterogeneous - preferences model (-0.04 to -0.29),
concluding that the two models might be upper and lower bounds on the estimates for
elasticity of demand for water. Also, they mention that these models are very costly to
estimate, since they require a large number of socio-demographic observations and have
complex non-linear functions.
25
Stone -Geary Form
A few authors used the Stone-Geary form to predict water demand and price elasticity
(Matinez-Espineira and Nauges 2004, Gaudin et al 2001, Al-Qunaibet et al 1985). The
function has already been used for food products, durable goods, transportation, and energy.
Gaudin et al.(2001) propose this form because it includes a quantity of water that does not
respond to price, allows elasticity to decrease as the price increases, and uses only two
parameters (y and P) for each good. y is defined as a threshold below which water
consumption is not affected by prices, while
P is
the preference variable. Basically, "The
consumer is faced with a given level of income and set of prices. The consumers first
purchases a minimum acceptable level of each good (the yj's) and then portions of each good,
for their leftover income, according to their preference parameter (the
Pj's)"
(Gaudin et al.
2001) Gaudin present the next form:
Q
= 7 +)
1* - P
-
YZ
where I and P are income and price. SGE (y,f)
are linear
PW
combinations of exogenous variables. So, the equation for non-constant y and P in the Gaudin
et al (2001) study is : Pw= (PO+P1C+P 2 SP+P 3APP) and yw = OCo+a 1C+cL 2 SP+cL3AAP
(yz was excluded from the model, as insignificant to the study)
Where C - days with rainfall; SP - Spanish population; AAP -average annual precipitation)
Gaudin et al (2001) found summer elasticities bigger than winter elasticities, and that more
than half of the water demand does not respond to price increase.
26
Table 3.5: Elasticities in studies that use the Stone-Geary Form
Author
Study area
Price
Income Elasticity
elasticity
Gaudin et al (2001)
Texas
0.19-0.28
Martinez-Espineira and Nauges (2004)
Spain
-0.1
0.1
Al-Quinaibet and Johnston (1985)
Kuwait
-0.77
0.211
Meta-analysis
Meta-analysis is the use of statistical techniques in a systematic review with a purpose
of integrating the results of the included studies. Espey et al. (1997) using meta-analysis
studied the factors that affect price elasticity estimates in recent studies in the USA. They
tried to explain differences in elasticity using differences in inclusion of variables in the
regression models. They found that long-run estimates are more elastic to than short-run
estimates; that the inclusion of income, population density, household size, temperature, and
seasonable variable do not influence the price elasticity even though they influence the
demand; also that evapotranspiration rates, pricing structure (increasing block rates were
found to be much more elastic), rainfall and the season influence the elasticity. Also, summer
elasticity was found to be bigger than winter elasticity.
Dalhuisen et al. (2003) in their meta-analysis study found that moderately high price
elasticities and reasonably low income elasticities are found in studies with increasing block
rates. Also, they find that the absolute magnitudes of price and income elasticities are greater
for areas with high income, that price elasticities in Europe are bigger than in the US and that
elasticities do not change with the date of the study, in other words they did not find
differences in elasticities of earlier and more recent studies.
27
3.17. Conclusion
Studies in water demand prediction and elasticity have come up with a wide range of
results. These studies have been conducted using different datasets, regression methods, price
increases and variables, that alter the results. Consequently, some correlation parameters have
been empirically proven. However, water demand and price elasticity are, no doubt,
influenced by local conditions and socio- economic variables. A consensus has not been
reached regarding the best methods to predict demand and elasticity. Most researches
conclude that more studies have to be done in water demand.
28
3.2 Water Metering
There are two main benefits of water metering. Water metering has been shown to
reduce demand, with a reduction rate generally from 10% to 35% (Lund, 1988). Second, with
water metering conservation is rewarded; households that use less water have a smaller water
bill.
There are several types of water metering schemes, depending on the pricing
specification. Universal metering is when meters are installed at all house connections. In an
optimal water metering plan, meters are only installed for consumers where the policy makers
conclude that installing a meter will increase the welfare; in other words were more water is
spent than the fixed water charge would cover
Prepaid Water Meters
Since one of the problems in Kuwait, is the inefficient collection of water bills, an
overview of prepaid water meters is presented.. The main advantage of prepaid water meters
is the efficient revenue collection. It is connected with the privatization of water utilities. The
World bank specifies that prepaid water meters" facilitate cost-recovery and accelerate
private sector participation in provision of water services"
(http://www.citizen.org/cmep/Water/humanright/meter/). The references about prepaid water
meters are mostly from the internet. Articles are divided into companies that advertise
prepaid mater meters and papers strongly opposing it. Prepaid water meters are used in areas
in South Africa, Philippines, China, Namibia, Swaziland, Tanzania, Brazil, and Nigeria
29
(http://www.citizen.org/cmep/Water/humanright/meter/). The non-profit organization "Public
Citizen" cites that in South Africa, people after being denied clean water went back to
collecting water from polluted sources which caused a cholera outbreak. Also, in the USA
they are used in areas without access to water infrastructure, and they were declared illegal in
UK in 1998."(Public Citizens).
ESI AFRICA (http://www.esi-africa.com/last/ESI 1 2003/031 18.htm) cites that:
"Already successfully installed in Mossel Bay, Ladismith and Kuruman as well as in a
number of trial sites in Cape Town, the Ecowater prepaid water meter has made significant
inroads into the decrease of outstanding water accounts in those municipalities.", and that
"Cape Town-based companies Rhomberg and Syntell (formerly Tellumat Electronics) claim
to have the answer to the country's problem of recovering hundreds of millions in outstanding
water revenues elsewhere".
The Company Hangzhou Jingda advertises a wireless remote water meter, specifying
that the water meter will automatically send its data to a concentrator every day at a certain
time or after consuming a certain quantity of water. Also, the valve will shut down when all
the water is used up, and will reopen when the consumer purchases water again. Water can be
purchased by using debit cards that are inserted into the water meter.
30
3.3 Water Usage
The use of water differs in various regions of the world, with a tendency that higher
industrialized countries use more water per capita than developing countries (Falkenmark and
Rockstrom, 2004). Figure 3.1 shows non-agricultural water usage by sectors for various
regions in the world.
Figure 3.1 NON-AGRICULTURAL USE OF WATER
USA
Household:
Service:
Industry:
20
210
@
0
@
H20
1
366 m3 cap' yr100
140
126
EUROPE
232 m3 cap 1 yr
Household:
Service:
Industry:
57
35
AFRICA
25 m3 cap- yr
Household:
10
Service:
8
Industr:
7
140
Falkenmark and Rockstrom, Fig. 3.2, p. 47.
31
1
Figure 3.2 (UNEP) presents a breakup of domestic water usage, using data from Swiss
households. It shows that 30% of water goes on toilets, while 20% of domestic water is used
for showers.
Figure 3.2
How much water do you use?
(data rfers to
swiss houscholder)
Activity
Toilet
Bath/shower
Washing machine
To oook, drink, to wash dishes (by hand)
Wash yourself and wash dresses (by hand)
Dishwashers
Other
TOTAL
Drinking and wash yourself and above all to
produce your food (if you are a good farmer
and you not waste your water)
water use (Vday)
4737
31.7
30,2
24.3
20.7
3.6
3.8
162
2000
The hidden water use
Conurgedity
water consumed (q
7
10
70
200
320
1000
1000
1000
4 000 to 10 000
1 litre of beer
1 litre of gasoline
1 cola
A single bath
1 kg of paper
1kg of bread
1 kg of potatoes
Television set
1 kg of meat
One pair of Jeans
&iurm DFEa
ProEtin
8 000
-AO:
Frshwa.te
Program
UN Environmental Protection Program
32
-
csmtiinn
up
Freshwater consumption in Europe
4. Price Proposal
Water Demand and Water Pricing in Kuwait
Abstract:
Kuwait is an arid country with limited natural water resources. Yet, water
consumption per capita is much higher than in other countries in the world and is estimated to
be around 450 L/capita/day. Certainly, one of the reasons is the fact that, even though water
has a pricing schedule in Kuwait, water bills are not collected efficiently; consequently there
is some amount of water that is being wasted. The main objective of this paper is to study the
potential impact of pricing as a tool for managing water demand in Kuwait. The original idea
was to construct a water demand model for Kuwait, but it was modified, because of the lack
of empirical data regarding household consumption characteristics and price influences on
demand. Instead water demand models described in the literature were adapted to Kuwait. A
pricing schedule is proposed, that consists of a free allowance, followed by a constant water
price. This proposal has the following logic: if water is consumed reasonably, only to satisfy
vital needs, it should be free. However, to limit over consumption, the quantity of water over
the allowance should be priced. Our results indicate that this pricing schedule would be
efficient in reducing demand significantly. The models results suggest that a price of water of
$1/m 3 , for water use in excess of a 150L/capita/day allowance, would reduce the demand by
about 35 percent (with a range of around 20-40 percent, depending on the demand model
used).
33
4.1. Introduction
Kuwait is an arid country, rich in oil but poor in natural water resources. The annual
rainfall is around 110 mm, and because of factors such as the high potential evaporation rate,
surface runoff and groundwater recharge from rainfall are rare (Fadlelmawla and Al-Otaibi,
2005).
The water resources in Kuwait are obtained from groundwater, desalination and
treated wastewater. The groundwater is mostly brackish and is not used in a sustainable way,
since the extraction from the wells is greater that the natural recharge to the aquifers. The
annual groundwater production in 1999 was around 118 million m3 , while under steady-state
conditions, the recharge, generated through lateral flow from Saudi Arabia, was estimated to
be 44 million m 3 /year (Fadlelmawla and Al-Otaibi, 2005).
Kuwait gets most of its potable water from desalination. Multi-Stage Flash is
currently the dominant desalting technology. The desalination capacity of Kuwait is 1.65
million m 3 /day (Hamoda, 2001), and the estimated cost to produce water is about 3 $/m3
(Darwish and Al-Najem, 2005). This cost was calculated based on a price of around
$40/barrel and an assumption that energy costs are equal to 45% of the total cost to desalinate
water.
Currently, more water is consumed in Kuwait than is necessary. The water demand in
Kuwait is around 453 L/capita/day (Darwish and Al-Najem, 2005), When the demand is
compared with other countries (USA - 333 L/day/capita; France 164 L/day/capita; Germany
34
127 L/day/capita (EWA 2005 yearbook)), a need to decrease water demand in Kuwait is
evident.
The main purpose of this study is to analyze the potential impact of pricing as a tool
for managing water demand in Kuwait. There may be several reasons for the large demand
(pipeline leakage, high temperature, etc.) but one important reason is certainly the low cost of
water to consumers. Water does have a pricing schedule in Kuwait; however in reality the
water bills are not collected. Darwish (2005) cited that the total income acquired by selling
455 million m3 (Kuwait's annual water production) was 86 million US$ in 2002. The annual
government subsidy was $715 M in 2003.
It is known from the literature and from experience of water utilities that water
consumption shows certain elasticity to price increase. The original idea was to construct a
water demand model for Kuwait in order to quantify the influence of pricing on water
consumption. However, assembling a model requires data regarding household consumption
characteristics and influences that price increases have on demand. This kind of data is not
available, since household water consumption is generally not metered and because there has
not been a price increase in recent years. Instead a study was carried out using water demand
models reported in the literature and by recalibrating them for Kuwait. Five models were
used based on studies in several arid regions: California, Tunis, Australia, Saudi Arabia, and
Spain.
A number of simulations analyzing the influences of various pricing schedules on
demand were performed including constant prices, block tariffs, and a free allowance
followed by various pricing schemes. In this paper, results regarding constant price schedules
with and without an initial allowance are presented.
35
A pricing schedule is proposed that consists of a free allowance (e.g. 150
L/capita/day), followed by constant water charge (Figure 4.1). This pricing schedule has two
parameters, the allowance and the constant price. It was concluded that this pricing schedule
would be acceptable for Kuwait, economically and politically, and that it would help to
eliminate some of the waste of water. Basically, if water is consumed reasonably, only to
satisfy vital needs, it would be free. However, consumption of water beyond what is judged
by society to be vital need should be priced.
Since countries in the Gulf region have similar socio-demographic characteristics and
low water costs, this study would have obtained similar results in some of the other countries
in the region.
Figure 4.1: Price Proposal - A free allowance followed by a constant price.
Block tariff schedules
3.5
3
1.5
15
0.5-
01[1
0
5
10
15
demand [m3/capita/month]
36
20
25
30
4.2. Literature review
Water demand reduction due to price increases has been extensively studied in the
literature.
Most of the demand models are regression models. Typically, the demand is
derived as a function of price variables and factors such as income, household characteristics,
and weather.
Which price variable to use has been the most common question in the literature. A
consensus has not been reached about whether to use the average price or the marginal price
combined with some other variables. A reasonable assumption, supported by some
researchers (Nieswiadomy 1992, Shin 1985), would be that if the consumers think the water
bill is significant, they will put in the effort to learn about the exact pricing schedule and their
exact consumption and hence would be influenced by the marginal price. Otherwise, where
the water bill represents a small percentage of income, the consumer will react to the average
price.
All reviewed studies find that household income is a significant variable that increases
demand. Also, household size is frequently used in the demand equation (Nieswiadomy 1992,
Renwick et al 1998, Dandy et al.1997). Variables such as lot sizes (Dandy et al. 1997,
Renwick et al. 1998, Lyman 1992, etc.), density of households, number of faucets (Renwick
et al. 1998), and age distribution (Lyman 1992) have also been used. A number of researchers
found that seasonal changes and climates influence demand. Summer demand was found to
be more elastic than winter demand (Lyman 1992, Dandy et al. 1997, Griffin and Chang).
Also, studies have found that outdoor water use is more elastic than indoor.
37
Renwick and Green (1998) showed that non-price policies (campaigns, restrictions,
rationing policies, etc.) have influence in decreasing demand. Nieswiadomy (1992) cites
experience from a campaign in Tucson which was successful in decreasing demand only for a
few years; after a few years consumption increased back to its previous level; he explains this
occurrence by the fact that actual water prices did not rise. Furthermore, Nieswiadomy
suggests that education programs will probably have more effect in water scarce regions,
because of the general awareness of water scarcity.
Abu Qdais and Nassay (2001) studied the impact of the change of the pricing policy
in Abu Dhabi City. Before the change, the water consumption was 636 L/capita/day.
However, after changing the price policy from a fixed charge to a constant price per unit
volume consumed, the demand decreased. The elasticity was found to be -0.1.
A few authors used the Stone-Geary form to predict water demand and price elasticity
(Matinez-Espineira and Nauges 2004, Gaudin et al. 2001, Al-Qunaibet et al.1985). The same
function has already been used for modeling demand of food products, durable goods,
transportation, and energy. Gaudin et al. (2001) proposed this form because it allows
elasticity to decrease as the price increases and uses only two parameters (y and
P) for
each
product. y is defined as a threshold below which water consumption is not affected by prices,
while
p is the preference variable.
The form also specifies a quantity of water that does not
respond to prices. Gaudin et al. specify the logic of the form in the following way: "The
consumer is faced with a given level of income and set of prices. The consumer first
purchases a minimum acceptable level of each good (the yj's) and then portions of each good,
38
for their leftover income, according to their preference parameter (the
(2001) presents the following demand function:
Qw
= y/, + 6 1*
P"
Pi's)".
Gaudin et al.
(4.1),
P
where the subscripts w and z indicate parameters pertaining to water and to other goods,
respectively, while I and P represent income and price.
Price elasticities from different studies found in the literature are presented in Table
4.1. As seen, price elasticities vary with in a wide range. This is due to the difference in
socio-economic characteristics of the study areas, different price schedules and price ranges,
consumption levels, climate, income, awareness of water scarcity, and many other factors that
can not all be accounted.
Espey et al. (1997) and Dalhuisen et al. (2003) studied factors that affect price
elasticities in demand models. Espey et al. evaluated 24 journal articles, while Dalhuisen et
al. examined 64 studies. Both researchers found that studies that are based on regions with
increasing block rate structures have bigger price elasticities. Espey et al. (1997) found that
the inclusion of evapotranspiration and rainfall results in less elastic demand, while the
inclusion of temperature, population density and household size does not affect the elasticity.
Dalhuisen et al. (2003) found regions with higher income to have larger price elasticities.
39
Table 4.1: Price elasticities from studies of residential water demand found in the literature.
Authors
Study area
Form
Howe and Linaweaver (1967)*
USA
Gibbs (1978)*
Miami, Florida
Price
Price
variable
Elasticity
AP
-0.23
-0.51
Foster and Beattie (1980)
Exponential
USA
AP
-0.35 to -0.76
Billings (1982)
Lin/Log
Tucson, Arizona
MP & D
-0.66/-0.56
Schefter and David (1985)*
Wisconsin
-0.12
Chicoine et al. (1986)*
Illinois
-0.71
Chicoine and Ramamurthy (1986)
Linear
Illinois
MP (AP)
-0.6 on MP
Nieswiadomy and Molina (1989)
Linear
Denton, Texas
MP & D
-0.86
Griffin and Chang (1990)
Linear
USA
AP
-0.16 to -0.37
Riazaiza (1991)
Logarithmic
Saudi Arabia
AP
-0.4 to -0.78
-0.10
Denmark
Hansen (1996)*
Renwick and Archibald (1997)
Linear
California
MP & D
-0.33
Hoglund (1997)
Linear
Sweden
MP & AP
-0.20 on AP
Dandi et al. (1997)
Linear
Australia
MP & D
-0.63 to -0.77
Renwick, Green, McCorkle
Logarithmic
California
MP & D
-0.16 to -0.21
Linear
France
AP
-0.22
(1998)
Nauges and Thomas (2000)
(&MP)
Ayadi et al.(2003)
Tunisia
Logarithmic
* Data from Nauges and Thomas (2000)
40
AP
-0.17
4.3. Input Assumptions, Demand Models Used and Adaptations made
First, the assumptions made about the input to the demand models are presented. Then for
every model, basic features are presented and the different adaptations made are explained.
Input assumptions made
As discussed in the previous section, a number of models assume that water demand
is a function of household income and household size. Based on income and household size,
consumers were divided into 40 groups in the following way.
Data regarding household income distribution in Kuwait was not available, so
assumptions were made regarding the distribution. The monthly household income was
calculated based on the distribution in the USA (U.S. Census Bureau), by dividing the
household income distribution in the US by the ratio of GDP per capita in the US and in
Kuwait. Non-Kuwaitis were assumed to have half of the GDP per capita as Kuwaitis, and
income distribution was calculated in the same way. GDP per capita values for the US,
Kuwaiti and Non-Kuwaiti households are presented in Table 4.2.
Table 4.2: GDP per capita in USA and for Kuwaiti and Non-Kuwaiti
GDP per capita
USA
Kuwaiti
$35,721
$21,300
Non-Kuwaiti
$10,274
US Census Bureau, 2004-2005 yearbook
World Factbook, 2005 (CIA Website)
Information was gathered from the U.S. Census Bureau for USA household income
distribution and the data was grouped into 5 income categories (Table 4.3). After dividing the
41
USA income by the ratio of GDP for Kuwaiti and non-Kuwaiti households to the US GDP
(ratio = 1.72 and ratio = 3.44, respectively), and the household income groups presented in
Table 4.4 were assumed for Kuwaiti and Non-Kuwaiti households and used as input in the
simulations.
Table 4.4: Assumed Household income
Table 4.3: Household income USA
Income(10uU$/month)
[70]
Under 1.25
15.8
1.25
2.92
25.6
2.92
6.25
36
6.25
12.5
17.9
Over 12.50
4.6
For Kuwaiti and non-Kuwaiti
I
Percent of
households in
each group[%]
15.8
25.6
36
17.9
4.6
Average income
Non-Kuwaiti
Kuwaiti
(1000$/month) (1000$/month)
0.26
0.51
0.60
1.21
1.33
2.66
2.72
5.44
5.56
11.13
Table 4.5 presents household size distribution data in Kuwait, acquired from the
Economic & Financial Quarterly of the National Bank of Kuwait (1999). This data was used
to incorporate the distribution of household size in to the demand models and was adjusted in
the following way.
The number of households in every category was transformed into the percentage of
the total number of households (Table 4.6). In table 4.5, the distribution of non-private nonKuwaiti households (households with members that are not related) is not presented. It was
assumed that this group has the same distribution as private non-Kuwaiti.
Every income group (specified previously) was assumed to have a household size
distribution computed from table 4.6., with respect to nationality. Therefore, following table
4.6, every income group (four Kuwaiti and four non-Kuwaiti) was divided according to
household size into four groups. This resulted in a total of 40 consumption groups that were
used in the simulations.
42
Table 4.5: Households in Kuwait
National bank of Kuwait (1999)
Type of
Household
Bachelor (one
member)
2-5 members
6-9 members
10+ members
Total private*
Kuwaiti
NonKuwaiti
Total
14,710
45,989
38,039
40,645
139,383
96,578
80,945
21,696
10,309
209,528
111,288
126,934
59,735
50,954
348,911
Non-private*
--
85,640
85,640
Grand Total
139,383
295,168
434,551
Table 4.6: Household size
distribution used in the simulations
HH
Kuwaiti Non-Kuwaiti
1
46%
11%
2-5
39%
33%
6-9
10%
27%
>12
5%
29%
Models Used
Models from five continents were used in these simulations. Figure 4.2 shows the locations
where the models were estimated
I California Model
1 Tunis Model
San~di Arahia Model
Snain Model
Australia Model
43
Saudi Arabia Model
Specification
Rizaiza (1991) studied water consumption in four major cities in Saudi Arabia:
Jeddah, Makkah, Madinah, and Taif. The study was based on data collected from a socioeconomic survey conducted in 1985, water and sewage department circulars from Saudi
Arabia, and from other publications. Residents that are supplied from the public water
network and tankers were analyzed. Since the public network is the subject of this study, only
that part of the paper is reviewed. The price elasticities, for areas connected to the public
network, were found to be -0.78 to 0.06. The average demand of water was 350 L/capita/day
with a price of 0.08 $/m3 for the first 100 m3 /month. The model parameters were estimated
using ordinary least square framework.
The water demand equation uses a logarithmic functional form, and calculates annual
water demand per household as a function of a constant (city dependent), income, average
price, family size, temperature and garden possession according to the following form:
Log(Q) = ao +a, -Log(INC) + a 2 -Log(PRIC) + a3 -Log(FSIZE) + a 4 -Log(TEMP) + a, -GRDN (4.2)
Where:
Q
= annual household demand; INC = household income, PRIC = average price;
FSIZE = family size; TEMP = temperature; GRDN =1 if family owns a garden, GRDN = 0
otherwise. In table 4.7 the coefficients for this model are presented.
44
Table 4.7: Coefficients used in the Saudi Arabia model
Coefficient
Parameter
0.1
GARDEN
-0.006
MAKKAH [city]
-0.16
MADI [city]
0.847
TAIF [city]
-0.78
LOG(PRICE)
0.44
LOG(FSIZE)
0.090
LOG(INCM)
1.26
LOG(TEMP)
Adaptation
Using the water demand equation (4.2) and coefficients (Table 4.7), demand was
computed for every consumption group using different prices. Average temperature was
assumed to be 31 C; the Garden dummy was not used, since data about garden ownership was
not available.
In order to simulate the effect of a free monthly water allowance in the pricing
schedule, the first 4.5 m 3/capita/month for every consumption group was computed with a
price of $0.1/M
3
(since the model has a logarithmic from, a zero price could not be
calculated), assumed to be close enough to free water. All consumption over the 4.5
m 3/capita/month was priced at the constant price rate.
Next, the model coefficients were adjusted to Kuwait in the following way. Since
there has not been a price increase in the recent years, the only point that is known on the
demand-price graph is a value around 450 L/capita/day for a situation when water is almost
free. Demand was computed to get a value of around 450 L/capita/day for a $0. 1/m3 price of
water. The model was modified for this point by changing the constant (city dependent) and a
value of zero was obtained.
45
California Model
Specification
This model, developed by Renwick and Green (1998), was based on data in California
for about 7.1 million people from eight water agencies during 1986-96. The model focuses on
the influence that price policies and non-price policies have on decreasing water demand.
Three kinds of equations are used in this study: water demand equation, price equations and
climate equations. Household water demand was derived as a function of price variables,
household income, lot size, precipitation, temperature, and non-price policies and has a
logarithmic functional form.
Water Demand Equation:
in W, =3o + AJ -1n(Mj )+8 2 -In(Di,+ AJ -ln(INC) ++L fl -(NPDSM)+,
3
-in(PREC,±+ A
4
-LOT
i=4
(4.3)
i=l,...,8 agencies (cities), t = 1,...,96 months (time)
Wit = Household Water Demand per month
MPit = Marginal price
Dit =Difference variable (the difference between what would have been paid if all units were
purchased at MP and the amount paid under the block pricing schedule)
INCit
HHit
=
=
Income in $1000
Number of household members
NP DSM = 8 non-price Demand Side Management (DSM) policies
PREC = precipitation
46
Values of some of the coefficients for the water demand equation of the California
Model are presented in the column named "California" in Table 4.8. Price and climate
equations are not presented here, since they are not related to this study.
Table 4.8: Coefficient values for adopting the California Model
California
Kuwait
Coefficient
Description
value
value
p0
Interception
2.61
2.35
01
D2
MP
Difference variable
Income
Household number
-0.16
-0.01
0.25
0.18
-0.16
-0.01
0.5
0.2
P3
p4
In this study a range in marginal prices of $0.16 - $1.6 per m3 was analyzed. The
model incorporates: price policies, alternative non-pricing campaigns, and seasonal and
climatic variability on demand. The parameters of equation (4.3) were estimated in a
generalized least-squares framework. The model estimates 16-20% price elasticity. The
authors find that non-price policies have influence on demand. Also, the authors conclude
that "price policy may achieve a larger reduction in aggregate demand in lower income
communities than in higher income communities".
Adaptation
Since the household allowance was specified to be per capita, that is a function of household
size, a household size variable was needed in the model. However, in the original model the
household size is not directly incorporated in the demand equation, but it is part of the price
equation.
One of the authors of the "California" model, (Renwick) published a similar paper
with Archibald (1998) that was based on data also from California, during the same period.
47
The household size variable was used in the demand model of this paper. However this model
(Renwick and Archibald 1998) was not used for this study, since it has a linear form, and
coefficients were found to be harder to calibrate. Since a household variable was used for a
similar study, the variable was added to the demand equation. A similar value of the
coefficient from the price equation was used for the demand equation for household size. In
the simulations, the household size did not have significant influence on the price elasticities;
it was used to be consistent with the price proposal.
Using the water demand equation (4.3) and coefficients of Table 4.8, demand was
computed for every consumption group for a range of prices. Then, the model coefficients
were recalibrated for Kuwait. In order to simulate the effect of a free monthly per capita
water allowance in the pricing schedule, the first 4.5 m 3/capita/month for every consumption
group was computed with a price of $0.1/M 3 , assumed to be close enough to free water. All
consumption after that quantity was priced at the constant rate. The income coefficient was
increased by a factor of 2, since the GDP in the USA is close to twice the GDP per capita for
Kuwaiti and Non-Kuwaiti households (possibly this assumption is not adequate, since the
function has a logarithmic form). The constant coefficient was then recalibrated to get a
3
demand of around 450 L/capita/day for a price of $0.1/M . Other coefficients remained the
same.
Table 4.8 (previous page) shows the values of recalibrated coefficients (Kuwait
column)
48
Australia Model
Specification
This study by Dandy et al. (1997) differs from previous research studies because the
influence of a free allowance in the pricing regime is analyzed. It was based on data from the
metropolitan area of Adelaide, Australia. The results of this study showed that: "consumption
above the allowance is more sensitive to income (or property value), climate variables
(summer moisture deficit and winter evaporation), and pool ownership than consumption
below the allowance but responds to the need of water as determined by plot size, household
size, and number of rooms no differently from consumption below the allowance" (Dandy et
al. 1997). There are two variations of the model, static and dynamic. The dynamic model
includes lagged water consumption. The two variations also differ because different
coefficients are used in the demand equations. The models are linear and are applied on an
annual time scale. Also, models with climate variables are included in this article.
The demand model is specified differently for consumers that are above and below
the allowance. For consumers below the allowance, water demand is a function of lagged
consumption Q-1, property value, and variables such as household size, climate, etc. (Z
variables) and Dy (dummy variable for year 1992).
D = 0 for
Q < A:
Q=aO +al -Q_ +/A I+BZ+O.D +u (4.4)
While for consumers above the allowance in addition to the previous variables, demand is
also a function of price (MP and D) variables, with different coefficients used, as seen in
equation (4.5).
49
D = 1 for
Q>A
Q = ao +
o +(a, + 91) -Q_ +(#A+y
1
1).I+(B+F).Z+6.D,
+ DP+u (4.5)
where:
Q = quantity of water
consumed
A = annual allowance
I = property value
P
=
a vector of price variables (marginal price, difference variable)
Z
=
a vector of other variables (household size, climate,, etc.)
D
=
variable showing if demand is above or under the allowance
Q-i= quantity of water consumed in the previous year
Dy=I for year =1992 and Dy=O, otherwise
In table 4.9 the coefficients used in the annual static consumption model (without lagged
consumption) are presented.
Table 4.9: Coefficients used in the Static Annual Australian Model
D=0
D= 1
Property
value
Plot
size
No.
residents
No.
Rooms
Pool
0.712
0.751
0.011
0.008
19.51
2.84
12.69
2.72
69.24
-56.4
Marginal
Price
Difference
variable
Constant
35.73
-404.4
-1.37
366.35
Adaptation
The original coefficients were used for the simulations (table 4.9), with the exception
of the following. Dy was not used since it is specific to Australia and the lagged consumption
was not included since the static annual model was used. Property value was not used in the
simulations, because of the lack of data.
The original model used property value instead of income, since not enough data was
available to the researchers regarding annual income in Adelaide. However, in the following
50
simulations income was used instead of the property value. The corresponding income
coefficient was computed so that the model would produce a demand of 450 L/capita/day for
a zero price of water in the simulation of a free allowance followed by a constant price. The
property value coefficient was divided by 133, to obtain the Kuwaiti demand for a zero price
of water. This would indicate that for an average family, the value of the property that they
own would be equal to the income that they earn in 133 months (11.1 years). In table 4.10 the
values of the income coefficients are presented. Also, Australian dollars were used in the
simulation, and then converted to US dollars.
Table 4.10: Income Coefficient values for adopting the Australia Model
possibility
value
D<A
D>A
0.00535
0.011
The Australia model is based on an annual allowance of 136 m3 per household. Since
the average household in Adelaide is 2.6 members/household (Australian Bureau of
Statistics, 1991); the allowance is in average 143 L/capita/day. This is close to an equivalent
to the allowance that will be proposed in this paper for Kuwait. However, the average
household size used in this paper for Kuwait is five members per household. Because of the
fact that the quantity of the allowance in this paper is defined to be per capita, this difference
in household size requires adjustments to be made to the model for Kuwait.
Water demand decrease due to price influences in this model is a function of the
marginal price (MP) and the difference variable (DV). The part of the demand equation (Eq
4.5) that computes the influence of the price has the formpA -MP +#82 -DV. In a pricing
schedule that consists of a free allowance followed by a constant price, the DV variable is
equal to the quantity of the household allowance multiplied by the marginal price of water,
51
with a minus sign. If the household allowance (AH) is specified to be per household member
with an allowance of 143 L/capita/day (= 52.2 m 3/capita/year), then:
.-MP+
-DV = -404.4 -MP -1.37 -(-AH ' MP)=
2 - DV = -404.4 -MP -1.37
=-404.4.MP+1.37-52.5 -HH -MP
(4.6)
where:
MP - marginal price; DV - difference variable; P1, P2 coefficients from table 4.10
AH - annual household allowance; HH - household size
For a household size of 2.6 member/household (Adelaide), equation 4.6 is equal to
-218.08 -MP. However if the same logic is applied for a household size of 5, equation (4.6)
becomes equal to -142.8 -MP, resulting in smaller price elasticity, just because of the size of
the household; these two factors should not be correlated. For a household with 12 members,
the price variables would be equal to 453.70 -MP, implying that an increase in prices would
increase the consumption for big households.
The model was adopted for the household size in Kuwait in the following way. The
1.37 was decomposed: 1.37 = 0.527 -HH (for HH = 2.6 in Adelaide), so consequently:
p, -MP+p/2
DV=MP.(-4044+1.37-52.2-HH)=M(-4044+0.527-HH-APM) (4.7),
where APM is annual water allowance per household member and 52.2 m 3/capita/year =
0.143 m3/ capita/day
The allowance had to be constant in the model; the variations in household sizes
would influence the price elasticities. For this model a fixed allowance of 275 m 3 /year (or
0.75 m 3/household/day) was used. This amount was chosen because in the model, the average
household size is 5 members per household, resulting in an average of 150L/capita/day. So a
52
value for
P2 =
-0.527 was used in the simulations. In addition, the household size variable has
a much smaller influence than the water price has on demand in the model.
Tunis Model
Specification
This model, developed by Ayadi et al. (2003), was based on water demand data collected in
Tunis 1980-1996. The authors first divided the consumers into five brackets based on water
demand. Based on similarity in changes in demand for the observed period, the lower two
brackets were combined into the lower block, and the higher two brackets were combined
into the higher consumption block. The model does not use the middle bracket.
Different coefficients were used for the high and low demand blocks, specifying that
the upper block has bigger price elasticity than the lower bracket. The water demand equation
(8) is a function of income, average price, network size, rainfall and quarterly dummies. The
model also computes household shifting from one bracket to the other (9).
Demand equation:
Log(C) = ao + a -Log(R) + a 2 - Log(P)+ a, -Log(N) + a 4 - Log(RL)+ Z a, QD,, (8)
s=1,2,4
Portion of households in each bracket:
Log(NB / N) = yo + Yj -Log(P)+72 - Log(N) +
-Log(RL)
+ E 7s -QD, (9)
s=1,2,4
Where:
C= average consumption of water per household [m 3/month]
R = average monthly income of households ($1000)
P = average price paid by household ($/month)
53
N = network size incorporated to capture the effect of network expansion
RL= indicator of rainfall
QD = quarterly dummy
NB = number of consumers in each bracket
Tables 4.11 and 4.12 represent the coefficients used in this model for the Greater Tunis area.
Table 4.11: Water Demand Equation coefficients for the Tunis model
Higher block
8.65
0.05
-0.34
Network size
Lower
Block
3.1
0.06
-0.08
0.02
Rainfall indicator
-0.1
-0.1
Coefficients
aO
aI
description
interception
Income
Price
a2
a3
a4
-0.16
Table 4.12: Consumers Proportion equation for the Tunis model
coefficient
description
YO
12
intersection
Lower Block
3.27
Higher block
9.84
price
0.05
-0.62
number of h
0.03
-0.21
Adaptation
The rainfall, quarterly dummies and network expansion variables were not used in the
simulations. The first and second Kuwaiti income groups (table 4.4) compose the lower
bracket and the fourth and fifth Kuwaiti income groups (table 4.4) create the upper bracket.
This led to a smaller percentage of the total population in the lower bracket, and a larger
percentage in the higher bracket, which consequently increases the price elasticity, because
the model is specified to have a bigger elasticity for the upper block. The u0 coefficients were
calibrated to get a demand similar to Kuwait (about 450 L/capita/day for a situation when the
54
water price is near zero). The 7o coefficients were changed in order to match the group
breakups specified in the first section (Table 4.4). Tables 4.13 presents the coefficients that
were changed in adopting the Tunis model.
Table 4.13: Changed coefficients for adopting the Tunis Model
Coefficient
ao
70
Kuwait
Higher block
Lower Block
1.8
1.25
2.035
1.525
description
interception
intersection
Tunis
Higher block
Lower Block
8.65
3.1
9.84
3.27
Spain Model
Specification
From the three models that are present in the literature that use the Stone-Geary functional
form the model developed in Spain was selected since Spain has a GDP similar to that of
Kuwait, and because the MP and difference variables were used in the model, which should
be a better fit for the price schedule proposed. The model was estimated in Feasible
Generalized Least Squares framework (FGLS).
The basic form is:
Q
=(1 -
)
+8 1'-+a,-BAN, +a 2 - POP, (4.10)
Pt
where:
Q, = average per capita consumption
Pt= the marginal price of water
It= virtual income, the difference between the average salaries and the difference variable
BANt = binary variable, indicating influence of out-door-use bans
POPt
=
daily hours of supply restrictions
y = the minimum consumption level (not affected by prices)
55
P
= "the marginal budget share allocated to the good considered"
Coefficients used in this model are presented in table 4.14.
Table 4.14: Coefficients used in the Stone-Geary model
Value
Description
Coefficient
0.008
Preference variable
0
4.7
Income*
ai
Ban on outdoor use
-0.125
a2
Restrictions
-0.05
Adaptation
The BAN and restriction variables were not used. The same values of the coefficients (yo and
p,) from the original model were used in simulating demand in Kuwait. Aggregate data was
used in the simulation; for income the GDP per capita for Kuwait was used ($20,547).
Model constrains and assumptions
The situation in Kuwait, were water bills are not collected efficiently differs from the
situations in the countries were the original demand models were estimated. A clear answer
about what is the difference between a free water situation and Kuwait's water pricing
situation (water does have a pricing schedule, but bills are not collected efficiently) and how
this might influence the water demand can not be found in the water demand literature. Also,
in the studies used in the paper, water prices increased initially from a certain price level
3
significantly bigger than a price of zero. Also, simulating a price increase of $2/M (e.g. from
0 to 2$/M 3 ), is a bigger price range than most of the original models were estimated in.
56
Also, in lack of data not all of the variables that were specified in the original models
were used in simulating water demand in Kuwait. Weather, density of housing, seasonal
influences and other variables were assumed constant and incorporated into the constant
(intersection); this might have an influence on the demand reduction.
In models that use the marginal price numerical errors occur when due to a price
increase, a consumption group shifts from one water price in the block tariff to another. So,
when the price of water gradually increases, the water demand of a group gradually
decreases. When the demand comes near the border of the blocks tariff (e.g. allowanceconstant price), it does not fit, either above the allowance, either under the allowance. In
other words, when demand is computed for the lower block (with the MP for the lower block,
and therefore a less negative price variable), it exceeds the lower block demand, going into
the higher tariff. But, when computed for the higher block (with a higher MP and a more
negative price variable), it does not reach the higher block. This numerical problem was
solved by putting the consumption group on the border between blocks.
4.5. Results
4.51 Simulation of a Constant Price Schedule
Results regarding the influence of a constant price schedule on per capita demand are
presented in Figure 4.3 and show that pricing influences the demand. All models compute
different price elasticities, since they are based on data from different studies. The models
that have a logarithmic or Stone-Geary functional form show that after a price of around
57
$0.8/m 3 , further price increase does not have influence on demand, proposing that there is a
quantity of water that is not influenced by the price.
Government subsidy is shown in figure 4.4. It was simulated for a cost to produce
water at $3/M 3. The subsidy was computed for a population of 2.2 million, to be consistent
with Kuwait's current population. All models show that water pricing would significantly
decrease monthly subsidy by around 60 million USD, for a price of $1/M3.
Figure 4.3: Decreases in demand for selected models for a constant price schedule without an
allowance.
Decreases in Demand for Selected Models
0.55
0.50
CL
0.45
0.40
-.-
0.35
0.
10
V
Australia
0.25
0.20
0.15
0
Tunis
0.10
0.05
o .n
0.0
0.2
0.4
0.6
0.8
1. 0
constant price [$/m 3]
58
1.2
1.4
-A-
California
Saudi Arabia
-w-
Stone-Geary
Figure 4.4: Monthly government Subsidy in Kuwait for a constant price schedule without an
allowance.
Monthly Government Subsidy
120--
100
Tunis
-u-Australia
California
80
-
Saudi Arabia
-m-Stone-Geary
_
60-
(Spain)
40
20-
0 0.0
0.2
0.4
0.6
0.8
1.0
1.2
1.4
1.6
1.8
2.0
constant price [$/m 3
4.52. Simulation of a constant price following a free allowance
As specified earlier, a simple pricing schedule is proposed: a constant price schedule
with a free initial allowance (Figure 4.1). It is specified by two parameters, the quantity of the
allowance and the constant price per m3 . To specify the quantity of the allowance water
demand was in Kuwait was compared to other countries. Most European countries use around
150 L/capita/day. This amount satisfies the need of an average person. Since the goal of any
pricing schedule is to eliminate the waste of water, all water above the allowance needs to be
priced.
The first 4.5 m 3/capita/month (150 L/capita/day) would be free of charge, while all
consumption afterward would be charged by prepaid water cards. Hence, at the beginning of
every month the prepaid card would be reset for the meter to accept a new amount of water.
59
In Figure 4.5, the influence of pricing on demand is presented, indicating that this kind
of pricing schedule will decrease demand. Similar to the previous simulation, the Log models
and the Stone-Geary model compute that after a price of 0.8 $/m3 the demand does not show
much reduction with further increases of the price (Figure 4.4), indicating that there is a
quantity of water that is not elastic to prices, while the linear Australian model suggests that a
price of $1.2/M 3 would decrease the demand to a consumption level of around 200-300
L/capita/day, depending on the model used (Figure 4.5). So, the models show that a price of
water of $ 1/M 3 , after a 150L/capita/day allowance, would decrease the demand by 25%-55%
percent, depending on the demand model used, with an arithmetic average of around 40%. The
circle on Figure 4 represents the point on the demand-price graph for which the coefficients
were recalibrated in the models
Government Subsidy (Figure 4.5) was computed in the same manner as in the
simulation without an allowance. All models computed a similar decrease in subsidy, of 40
million dollars per month, for a price of 1$/M 3.
60
Figure 4.5: Decreases in demand for selected models for an allowance followed by a
constant price schedule.
Decreases in Demand for Selected Models
0.550
0.500
0.450
5z
0.400
__California
0.400
0.350
L
EI
0.300
-= unis
-*--Californi
-- Australia
0.250
u-T-ni
.Stone-Geary
0.200
-)-
0.350
Saudi Arabia
0.150
0.100
0.0
0.2
0.4
0.6
0.8
1.4
1.2
1.0
1.6
2.0
1.8
constant price [$/m 3]
Figure 4.6: Monthly government Subsidy in Kuwait for an allowance followed by a
constant price schedule
Monthly Government Subsidy
120.0
100.0
80.0
-+-
60.0
--
0
E
40.0 -20.0
0.0
0.0
0.2
0.4
0.6
0.8
1.0
1.2
constant price [$/m3]
61
1.4
1.6
1.8
2.0
California
-- Tunis
Stone-Geary
-Australia
Saudi Arabia
4.53. Elasticities
Models used in these simulations are based on studies from arid regions in five
continents: North America (California), Europe (Spain), Asia (Saudi Arabia), Africa (Tunis)
and Australia. Since the models are based on data sets from different regions of the world, they
predict different price elasticities. Table 4.15 presents the elasticities for marginal (constant)
prices in the original model and those found in the simulations. As can be perceived in table
4.15, elasticities in the simulations differ from original values for some of the models. This
divergence might in part be due to the fact that all variables were not put into the models.
However, the allowance also makes a significant impact on the elasticities.
Table 4.15: Elasticities obtained in the simulations and in the original models
Model
At price
Const. prices
California
Australia
Tunisia
Saudi Arabia
[$/m 3]
0.65
0.73
0.8
0.75
-0.16
-0.89
-0.41
-0.86
price
Original
model
-0.13
-0.33
-0.33
-0.35 (AP)
-0.16 to -0.2
-0.63 to -0.77
-0.17
-0.78
Allowance & const
-0.78(MP)
Stone-Geary-Spain
1
-0.24
-0.24
-0.1
California model: For the simulations without an allowance, the elasticity is in the range of the
original model. However, in the simulations with the allowance, the elasticity is smaller. This
is partly due to the fact that when an allowance is simulated, a fraction of consumers are
consuming water below the allowance quantity, getting water for free, and are not influenced
by price increases.
Australia model: The model was based on an allowance; it specifies that a portion of the
consumers consume within the allowance, and are not influenced by prices. So, when
62
simulations without an allowance were performed, all the consumers are influenced by price
increases, resulting in a bigger elasticity. Simulations with an allowance computed smaller
price elasticities, mostly because the allowance in the simulations was bigger than in the
original model. Also, the Australia model has a linear form, so elasticities differ with prices.
For the California and Australia Model, income groups that numerically did not fit either above
or below the allowance were fixed to be on the border. To check if this would influence the
elasticities, a simulation was conducted using only groups that do not have these problems at
any price ranges (California-33 groups Australia-27 groups). The results showed almost
identical elasticities; the California model showed that the way the shifting between blocks was
computed has an insignificant influence on elasticity. For the Australia model, the elasticities
were the same; the demand function was totally linear when only groups without errors were
used, while when all groups were used there was some non-linearity (Figure 4.4).
Tunis model, the assumed household income distribution (table 4.4) led to a smaller percentage
of the total population in the lower consumer block, and a larger percentage in the higher block
than in the original model, which consequently increases the price elasticity, because the model
is specified to have a bigger elasticity for the upper block.
Computing marginal price elasticities for models that use the average price as a variable might
not present the results in the best way; however consistency was the governing principle.
63
4.54 Influence of the allowance and price on demand reduction
The influence of the allowance on demand reduction for various prices is presented in figure 4.7
for the Australia model and in figure 4.8 for the Saudi Arabia model. In the simulations, the
initial demand was computed for prices equal to $0.25/M 3 for the Australian and Saudi Arabia
models and reduction percentage was based on the corresponding demand. Both models showed
similar results indicating that the price has a bigger influence on demand than the allowance.
Also, for different values of the allowance, the initial values of demand slightly differ; however,
these initial values have insignificant influence on the demand reduction percentage.
Table 4.16 shows demand reduction in the simulations for all models at a marginal price of
$1 /m 3 for an allowance of 150 L/capita/day.
Table 4.16 - Reduction in demand in simulations at the marginal price of 1$/m 3
Model
California
At the marginal pricing of 1$/m3
With allowance of
Reduction from an initial
150 L/capita/day
Without allowance
price of [$/m3]
17%
19%
0.25
Tunisia
0.25
45%
40%
Saudi Arabia
0.25
66%
41%
Australia
0.25
57%
28%
Stone-Geary Spain
0.25
41%
41%
45%
34%
Average
64
Figure 4.7 - Influence of the allowance and price on demand reduction - Australian Model.
Influence of the Allowance and Price on Demand for the Australian
Model
2.25 2.00
-
1.75 A
0.7
0.50
0.25
75
100
150
125
175
200
allowance [Ucapita/day]
Lines represent areas of equal reduction percentage from a demand computed for a price of $0.25/m3 .
Figure 4.8- Influence of the allowance and price on demand reduction - Saudi Arabia Model.
Influence of Allowance and Price on Demand Reduction for
the Saudi Arabia Model
2 2o
x
1.751.5 -
E
1.25-
0.75 -
0.5
0.25
75
100
125
150
Allowance [Llcapitalday]
175
200
Lines re resent areas of equal reduction percentage from a demand computed for a price of
0.25$/m .
65
225
4.55 Changes in the distribution of the consumption
The distribution of computed water demand, using the Australia model for simulations of a free
allowance followed by a constant price, for various prices are presented in Figures 4.9-4.11. The
figure shows that at a zero price of water, there will be no households with consumption less than
150 L/capita/day. However, for a price of $2/M 3, over 50% will consume less than that value.
Figure 4.9
Computed Distribution of Water Demand for a Free Water Situation
60%
50%40%
30%
20%
10%0%0-0.15
0.15-0.25 0.25-0.35 0.35-0.45 0.45-0.55 0.55-0.65 0.65-0.75 0.75-0.85 0.85-0.95
[m 3/capitalday]
66
<0.95
Figure 4.10
Computed Distribution of Water Demand for a price of 1$/m 3
60% 50%
40%
30% 20% -
10%
0% 0-0.15
0.15-0.25 0.25-0.35 0.35-0.45 0.45-0.55 0.55-0.65 0.65-0.75 0.75-0.85 0.85-0.95
<0.95
[m 3/capita/day]
Figure 4.11
Computed Distribution of Water Demand for a price of 2$/m 3
60%
50%
40%
30%
20%
10%
0%
-I
0-0.15
0.15-0.25
0.25-0.35
0.35-0.45
0.45-0.55
[m 3/capita/month]
67
~~'
F
0.55-0.65
0.65-0.75
0.75-0.85 0.85-0.95
<0.95
4.5. Conclusion
The results showed that pricing water in Kuwait would decrease the demand to an acceptable
level. For the simulation of a constant price with an initial free allowance, the models computed
that a $1/m
3
price of water, after a 150 L/capita/day allowance, would decrease the demand by
20-40 percent, depending on the demand model used, with an arithmetic average of around 35
percent. For simulations without an allowance, the arithmetic average of the demand reduction
for the five models was found to be around 45 percent.
Logically, marginal price elasticities were shown to be slightly larger for a schedule
without an allowance. However, paying for water in Kuwait is not the normal practice. So, water
bills could be perceived as an unnecessary toll that possibly would not be widely accepted. From
a political/socio-economic point of view, a free allowance would certainly be more acceptable
solution to address Kuwait's immense water demand problem. It would be for the Kuwaiti
government to quantify the two parameters, the price and the allowance, by selecting the most
appropriate values.
Since the main objective of charging water after an allowance is to eliminate the waste of
water, this kind of pricing would help in addressing this problem. Also, the results presented
would be similar for other countries in the Gulf region, where water is generally under priced.
The assumptions abut to the demand models to input should not influence the projected
demand reduction. However, since data with actual Kuwait demand characteristics is not
available, possible divergence might occur. But, since the models generally have similar results,
range of predictions of the influence of the proposed pricing schedule should be accurate enough.
More research has to be done in the water demand modeling for Kuwait. Research has to be
68
conducted to see what the water is spent on. Water needs to be metered. Data regarding
household consumption characteristics need to be collected, including household income,
household size, household water consumption, pool ownership, lot sizes, and landscape
irrigation.
69
5. Simulations of the Influences of Various Pricing Schemes
Prior to the price proposal, other simulations were conducted to see the influence of
various pricing schemes on demand. Simulations of the influences of block tariffs and the
influence of the percent of household income paid for the water bill were performed. The input
assumptions, consumer groups and variables used are the same as in the price schedule proposal
of chapter 4. Coefficients used slightly differ from the coefficients in the price schedule proposal,
since these simulations had the goal to study the demand models and were conducted in a
preliminary phase of this study. This chapter is presented after the price schedule proposal
chapter, even though they were conducted before it. In this section only the California and Tunis
models were used. The main results of these simulations are that an increasing block tariff
schedule will decrease demand more than a constant pricing schedule. Also, simulations showed
that a pricing schedule that would result in 1.5% of the total income to be paid for water by an
average consumption group would make the demand decrease to a standard consumption level.
70
5.1 Influence of water pricing and block tariffs on Demand and Government
Subsidy
5.11 California Model
Adaptation
The coefficients in the model were adapted to get a value of 430 L/capita/day for a price
of water of $0.75/M 3 . The next changes were made. The MP coefficient was increased assuming
that because of the lower pay in Kuwait, consumers would show a slightly larger influence on
pricing. The income coefficient was increased by a factor of 2.6*1.27 where 2.6 is the GDP ratio
for the USA and combined Kuwaitis and Non-Kuwaitis. 1.27 represents the parameters that were
not put in the model but are a function of income (lot size...). Also a cost to produce water of
$3/M3
was assumed. Table 5.1 summarizes the coefficients used in the simulations of block
tariffs. The problem of shifting between blocks (explained in the price proposal chapter) was
addressed differently than in the price proposal. It was addressed by always taking the lower
block. The errors made by this situation increase as the price increases, and can at higher prices
influence the demand by 10-15%. But since it was always calculated in the same way, it should
have insignificant influence.
Table 5.1. Original coefficients and coefficients used in the
California Model Block Tariff and Prohibition simulations
Coeff.
P0
P1
Description
Interception
MP
California
value
2.61
-0.16
P2
Difference variable
0.01
-0.01
P3
Income
Household number
0.25
0.2
0.85
0.24
P4
71
Kuwait
value
2.48
-0.256
Results
To show the influence that block tariffs have on water demand, three different tariffs
were used. One, with a const price of water, and two with different slopes of marginal price
increase (table 5.2 & Figure 5.1).
Table 5.2 Block Tariffs used in the California model
Demand [m3]
No
to
from
1
2
3
4
5
const.price
Blocki
Block2
$/m3
$/m3
0.72
0.74
0.76
0.81
0.85
$/m3
0.53
0.71
1.06
1.77
2.47
14
42
71
113
453
0
14
42
71
113
0.76
0.76
0.76
0.76
0.76
Figure 5.1 -Block pricing schedules used in the California simulations
Block pricing schedules
3.00
2.50
2.00
--.
E
(V)
1.50
Const. Price
-Block
--- Block I1
1.50
-6-Block2
1.00
0.50
0.00
0.00
0
I
I
I
20
40
60
80
I
I
I
100
120
140
water demand [m3]
Simulations were made for different prices and block tariffs. An average price schedule and a
block tariff were compared. Results show that changing the pricing schedule from a constant
72
price to a block pricing schedule will have similar influence on demand as a 50% increase in
price for a constant price schedule (Figure 5.2).
The influence that pricing has on demand for Kuwaiti (Figure 5.3) and non-Kuwaiti
(Figure 5.4) are presented. Since Kuwaitis have bigger average income, they consume more
water than Non-Kuwaitis. So, the decrease in demand for Kuwaitis would have a much bigger
influence on the total demand reduction than for Non-Kuwaitis. This suggests that a block tariff
schedule will be more efficient in decreasing demand.
The influences of the average price of water on total cost to produce water, total price
paid by consumers, and total subsidy by the government are presented, for a const price and the
third block tariff schedule in figure 5.5 for a total population of 1.9 million. Since block tariffs
compute smaller demand for the same average price of water, consequently they reduce the cost
to produce water and the total price paid by consumers. However, as the subsidy is the difference
between these two categories, the subsidy has a similar value for all pricing schedules. The
figure shows that pricing water an average rate of $1.5/M 3 , would reduce the subsidy by 30
million dollars (the cost to produce water was assumed to be $3 /m3 ).
Figure 5.6 shows the influence on demand for: 5 prices in block 3 (for MP & D), 5
income groups, 4 household size groups, and basic demand (constant). It shows that the income
and constant variables have the biggest influence in the demand equation.
73
11
-.-
- , -
=
-
,
-,-
-
-
-
MWFPR-
- -
-
-
--
---
-.-
Figure 5.2 - Influence of block tariffs for the California model
Influence on Total Demand in Kuwait
29.00
17.00
15.00-
0.75
1.00
1.25
average price
Figure 5.3
1.75
1.50
2.00
[$1m 3J
1900- Influence of pricing on Kuwaiti (California model)
Influence of block pricing on Kuwaiti
,..
105
O 100
E
2150
0
80
05
-U-Bloc
nuat
Inlec
price
-ver-a-Const
75
0.
7
E
u65
60
0.60
0.85
1.10
1.35
1rcn
1.60
average price [$/m3]
74
1.85
2.10
2.35
Figure 5.4 - Influence of pricing on Non- Kuwaiti (California model)
Influence of block pricing on Non - Kuwaiti
48
46
-
42
-4-Const. Price
-4-Block 1
- -Block 2
040-
-E
0 38
E 36
0
c 3432
30
0.60
0.80
1.00
1.40
1.20
1.60
1.80
average price [$/m3]
Figure 5.5_- Influence of Constant pricing and a block tariff system on the Total Cost, Price and
Government Subsidy for the California model
Influence of Water Pricing on Total Cost, Price, and Government
Subsidy
90
80
70
60
50
E
40
30
20
10
0
0.75
0.95
1.15
1.55
1.35
1.75
1.95
average price [$/m3]
Lower lines (red) - Block 2 pricing)
(Upper lines(blue) - Constant pricing
75
2.15
Figure 5.6 - California Model- influences of variables in the demand equation
Influence of the Components in the Sum of the
Demand Equation
for an average price of 1.9 $/m3
3. 002.
50
-O-B*ln(MP)
-U-B2*ln(D)
- Kuwaiti
-B3*ln(Inc)
B4I1n(HH)
00-. 50-
-4--Basic Demand
0.
--
2
-0.
3
o-
B3*ln(INC) -Non-Kuwaiti
4
50
-41.
00
50r
group-price
5.12 Tunis Model - Block Tariffs
Adaptation
Table 5.3 presents the coefficients used in the water demand equation in the Tunis simulations
for block tariffs. In this simulation the coefficients were calibrated to get a demand of
430L/capita/day for a price of 0.75 $/m3. The intersection coefficient was changed, while other
coefficients remained the same.
The Yo coefficients were changed in order to match the group break ups specified in the
first section. The first and second groups compose the lower bracket and the fourth and fifth
groups make the upper bracket. This led to a smaller percentage of the total population in the
lower bracket, and a larger % in the higher bracket; which consequently increases the price
76
elasticity, because the model is specified to have a bigger elasticity for the upper block. Values
are presented in table 5.4
Table 5.3 - Coefficients used in the Tunis model water demand equation for
block tariffs and prohibition simulations
Tunis
Kuwait
Coeff
cO
al
cL2
description
interception
Income
Price
Lower Block
1.5
0.06
-0.08
Higher block
2
0.05
-0.34
Higher block
8.65
0.05
-0.34
Lower Block
3.1
0.06
-0.08
Table 5.4 - Coefficients used in the Tunis model consumers proportion
equation for block tariffs and prohibition simulations
description
Kuwait
Higher block
Lower Block
Tunis
Lower Block Higher block
YO
intersection
1.454
2.462
3.27
9.84
Y1
Y2
price
number of h
0.05
0.03
-0.62
-0.21
0.05
0.03
-0.62
-0.21
Simulation
In order to compute the response of various blocks on water demand, the tariffs presented
in table 5.5 & Figure 5.7 were used in simulations.
Table 5.5 - Block tariffs used in the Tunis Model Simulations
Demand
Block1i Block2
[m3/household/monthl const. rc
3
$/m 3
$/m 3
to
$/m
from
No
1
0
14
0.76
0.72
0.6
2
14
42
0.76
0.74
0.72
3
42
71
0.76
0.76
0.8
4
71
113
0.76
0.81
1.3
5
113
453
0.76
0.85
2
77
Block3
3
$/m
0.53
0.71
1.06
1.77
2.47
Block4
$/m 3
0.01
0.53
2.12
2.82
3.53
Figure 5.7 Block Tariffs used in the Tunis Simulations
Block pricing schedules
4.00
3.50
3.00
4-
2.50
---
-
E
Const. Price
Block 1
-a- Block3
2.00
- Block4
C- 1.50
-*-
Block 2
1.00
0.50
0.00
0
20
40
100
60
80
water demand [m3]
120
140
Figure 5.8 shows the influence that block tariffs have on demand. A more increasing tariff
lowers the demand more. This is because an increasing block schedule influences the higher
consumers more, making the higher bracket have a higher average price.
78
Figure 5.8 - Influences of block tariffs on demand in the Tunis simulations
Influence of block tarriffs on demand
I
29.00
27.00--
E
25.00
23.00-
Constant price
-u-Block 1
Block 2
E
21.00
Block 3
0
19.00-
-.*0
-*-Block
4
I-
17.00
15.00
0.70
0.80
0.90
1.00
1.10
1.30
1.20
1.40
1.50
Average price [$/m3]
As can be seen on the next two Figures the higher bracket shows a significant decrease when the
average price paid by the bracket is increased (Figure 5.9), while the lower bracket only slightly
decreases (Figure 5.10). Figures are the same for all tariffs because water demand is computed
using the average price paid in considered bracket
79
Figure 5.9 - Responds of the Higher Bracket in the Tunis Model
Respond of the Higher Bracket to Water Prices and Block
Tariffs
120.00
115.00
110.00
-V
-+
105.00
V-
100.00
C
0-*E
95.V0
95.00
Constant price
Block 1
Block 2
Block 3
Block 4
90.00
85.00
80.00
0.00
0.50
1.00
1.50
2.00
2.50
3.00
average price paid by higher bracket[$/m3]
Figure5. 10 - Responds of the Lower Bracket in the Tunis Model
Respond of the lower bracket to water prices and block tariffs
35.00
34.50
34.00
V,
33.50
:E
0
Constant price
-u-Block 1
Block 22
-+-
0
V
33.00
E 32.50 32.50
-*-
32.00
E
31.50
31.00
30.50
30.00
0.50
0.70
0.90
1.10
1.30
average price paid by lower bracket[$/m3]
80
1.50
Block 3
-Block 4
This Figure (5.11) shows the influence of both block pricing and price increase on the
total cost to produce water, total price paid by consumers, and total subsidy by the government.
As can be seen, the higher block schedule, decreases demand the most, influencing all
parameters (for the same average price paid by consumers).
Figure 5.12 shows shifting between blocks when prices increase. The increase of water
price will make consumers shift from the higher bracket to the middle one (not calculated), and
from the middle to the lower bracket.
Overall, the results for this model show that the Lower bracket demonstrates a small
decrease in demand as a result of a price increase of $1/m
3
(8%), while the higher bracket
expresses a significant reaction to a 1$/m 3 price raise (up to 25%). Price increasing makes
consumers from the higher bracket shift to the lower, therefore decreasing the total demand.
Figure 5.11 -The Total Cost and Price of Water and Government Subsidy in the Tunis model
Influence of prices and block tariffs on Cost, Price, And Government
Subsidy
-- +--C.P.- Price
90.00
80.00 -
g
'
--.-- C.P. -Coste
- - - - C.P. -Subsidy
u
B1 - Price
70.00
60.00
:-
-
_
---
50.00--
B1 -subsidy
B2 -price
0
E
40.00-
B2 -Cost
B2 -Subsidy
30.00 20.00 --
B 1 -Cost
- - -
B3-price
B3 -Cost
10.00 _
B3 -Subsidy
0.00
0.70
0.80
0.90
1.00
1.10
1.20
1.30
Average price [$/m3]
81
1.40
1.50
--
B4 -Price
B4 -Cost
B4 -Subsidy
Figure 5.12 - Shifting between block in the Tunis model
Consumer shifting from bracket to bracket with increase of price
50%
45%
*0
)40%
~ ~~
-
~
-+- C.P. [%] of
0-
consumers in lower
-
bracket
35%
--
30%
C.P. [5] of consumers
in higher bracket
E 25%
o
20%
B4 -[%] of consumers
in lower bracket
10%10%
B4 -[%] of consumers
0
in higher bracket
5%
0%
0.70
0.90
1.10
1.30
average price [$/m3]
82
1.50
5.2. Influence of the percent of household income
that is paid for the water bill on demand
California model adaptation
In order to get a reasonable model for Kuwait, some of the coefficients were altered in the
following way:
"
Overall, coefficients were adjusted to get a consumption of 375 L/day/capita with a water
price of $0.15 m3 (currently in Kuwait water bills are not efficiently collected; since the
model has a logarithmic form a zero price is not computable).
" The coefficients remained the same as in the previous simulations (Table 4.1), with an
exception of the intersection. It was calibrated to get a value of 375 L/capita/day for a
price of $0.15/M 3 . Also, a cost to produce water of $3/M
3
was assumed. Coefficients used
in this simulation are presented in table 5.6.
Table 5.6 - Coefficients used in the California Model simulations of
the influence of the percentage of the household bill on demand
Coeff.
p0
p
p2
p3
p4
Original
Kuwait
value
2.61
-0.16
0.01
0.25
0.2
value
2.1
-0.256
-0.01
0.85
0.24
Description
Interception
MP
Difference variable
Income
Household number
In this simulation, block tariffs were constructed as a function of the percent of the household
income that is spent on the water bill by an average household. This was analyzed to show how
much an increase in water charging would influence the average consumer's budget. Also, the
decrease in demand was shown, and government subsidy was calculated.
83
The following Figures represent:
" Block tariffs that will produce a water bill that will be equal to a certain percentage of
income for an average household (Figure 5.13)
" The decrease in water demand, as a function of percent of income that an average
household has to pay for water (Figure 5.14)
The government subsidy for various ratios of the water bill and income. (Figure 5.15)
*
Figure 5.13 - Block Tariffs constructed to have a certain percentage of the Total income
Block Tariffs
Constructed for the Average Water Bill
to have a certain %of Total Income
5.00
4.50
4.00
3.50
2% of income
-of income
-5-1.5%
-0-1.5% of income
-*-1 % of income
income
-~-1
of income
income
0.5% of
-"" -0.5%
R3.00 F~ 3.00
E
2.50
2.50
U
CL
2.00
1.50
1.00
0.50
0.00
0
10
20
30
70
60
50
40
demand [ m3/household/month]
84
80
90
1 00
Figure 5.14 - The percentage of income paid for water
Influence of the percent of income paid for water on demand
by the average household
0.45
0.40
S0.35-
0.30
E
0.25
0.20
0.15
0.00%
0.50%
1.50%
1.00%
2.00%
2.50%
% of income paid for water per month
Red (horizontal) lines represent min and max values
1% of income equals average price of 0.5 $/m3 112% of income equals 1.28 $/m3
Figure 5.15 - Government Subsidy for various income percentages
Influence of the percent of income paid for water on Government
Subsidy for a 1.5 $/m 3 production cost
80
70-
60
0
E 50-
2 40
30
20
10
0.00%
0.50%
1.50%
1.00%
%of income paid for water
85
2.00%
2.50%
6. Conclusion
In order to address to the immense water consumption problem in Kuwait, the water demand
literature was extensively studied, and evidence that water demand is influenced by prices was
found. Due to lack of data concerning influences of price increases and household consumption
characteristics, the original idea of constructing a water demand model for Kuwait was modified.
Water demand models for several arid regions were adapted from the literature and recalibrated
for Kuwait.
Simulations of various kinds of pricing schedules were conducted. Input assumptions
were made, however it was concluded that these assumptions have an insignificant influence on
the predictions. Results showed that increasing block pricing schedules would be the most
efficient in reducing the demand in Kuwait. However, this kind of pricing schedule would not be
the best fit for Kuwait, from a socio-economic point of view, since paying for water is not a
custom in Kuwait and increasing block tariffs might not be widely accepted. Instead a free
allowance, followed by a constant price schedule is proposed. Results showed that this kind of
pricing schedule would be sufficient in reducing the demand to a reasonable level and will
eliminate the waste of water. Also, simulations showed that a price schedule that would make the
water bill a very small portion of the household income (1-1.5%) would be beneficial in reducing
demand.
Since an actual demand model was not assembled, the range of the prediction is large,
(between 20-40 % of demand reduction, depending on the demand model used). Generally the
California and Australia model have similar results; they predict smaller elasticities and demand
reduction. It can be assumed these two models represent the lower boundary of the demand
86
reduction. While, the Tunis, Saudi Arabia and Spain model have very similar results, and
represent the upper boundary in the demand reduction.
Also, results presented here would be similar for other countries in the Gulf region, since
they have similar characteristics. Certainly, more research has to be done in the water demand
modeling for Kuwait.
7. References:
Mousa Abu-Arabi and K. Venkat Reddy, 2005
R&D Challenges for Further Cost Reductions in Desalination
MENREC2 Conference, Amman, Jordan
Agthe and Billings, 1996
"Water-Price Effect on Residential and Apartment Low-Flow Fixtures."
Journalof Water Resources Planningand Management/ January/February
Agthe and Billings, 1996
"Water-Price Influence on Apartment Complex Water Use."
Journalof Water Resources Planningand Management / January/February
Dr. Al-Sabeeh, 2001
Desalinizationcountryfocus Kuwait
Al-Qunabeit, Johnston, 1985
"Municipal Demand for Water in Kuwait, Methodological Issues and empirical results."
Water Resources Research 2 (4)
Arbues, Garcia-Valinas and Martinez-Espinera, 2003
"Estimation of residential water demand: a state-of-the-art- review."
Journalof Socio-Economics 32(1)
Ayadi, Krishnakura, and Matoussi; Tunis, 2003
87
"A panel data analysis of Water Demand in Presence of Nonlinear Progressive Tariffs."
The Australian Bureau of Statistics', 2001 Census
Adelaide Social Atlas
Billings and Agthe, 1980
"Price Elasticities for Water: A case of Increasing Block Tariffs."
Land Economics 56 (1)
Billings, 1982
"Specification of Price Rate Variables in Demand Models."
Land Economics 58(3)
Chambouleyron, 2004
"Optimal Water Metering and Pricing", Water resourceManagement 18
Charney and Woodard, 1984
A test of Consumer Demand Response to Water Prices: Comment
Land Economics 60(4)
Chicoine and Ramamurthy, 1986
Evidence on the Specification of Price in the Study of Domestic Water Demand
Land Economics 62(1)
Dandy, Nguyen, and Davies, Australia, 1997
"Estimating Residential Water Demand in the Presence of Free Allowances."
Land Economics 73(1)
Dalhuisen, Florax, de Groot, and Nijkamp, 2003
"Price and Income Elasticities of Residential Water Demand: A Meta-Analysis."
Land Economics 79(2)
Darwish et al, 2004
"Towards Sustainable energy in Seawater Desalting"
Darwish et al, 2005
Substitution of MSF Desalting plants with Absorption Air Conditioning on Co
Generation Power plants in Gulf Area
Darwish and Al-Najem, 2005
"The Water Problem in Kuwait" Desalination177
Darwish, Asfour, Al-Najem, 2002
"Energy consumption in equivalent work by different desalting methods: case study of
Kuwait" Desalination15283-92
88
Economic & Financial Quarterly i/1999
Kuwait National Bank
Espey M., Espey J., and Shaw, 1997
"Price Elasticity of Residential Demand for Water: A Meta-Analysis."
Water Resources Research 33(6)
European Water Association Yearbook, 2005
Fadelmawla and Al-Otaibi, 2005
"Analysis of the Water Resources Status in Kuwait" Water Resources Management 19
Falkenmark and Rockstrom, 2004
"Balancing Water for Humans and Nature"
Foster and Beattie, 1981
"On the specification of price in Studies of consumer Demand under Block price
Schedules. " Land Economics 57(4)
Foster and Beattie, 1979
"Urban Residential Demand for Water in the United States." Land Economics 55(1)
Gaudin, Griffin, and Sickles, 2001
"Demand Specification for Municipal Water Management: Evaluation of the StoneGeary Form." Land Economics 77(3)
Griffin and Chang, 1990
"Pretest Analysis of Water Demand in Thirty Communities"
Water Resources Research 26 (10)
Hamoda, 2001
"Desalination and water resource management in Kuwait" Desalination 138
Hewitt and Hanemann, 1995
"A Discrete/Continuous Choice Approach to Residential Water Demand under block rate
pricing" Land Economics 71 (May)
Lyman, 1992
"Peak and Off-Peak Residential Water demand"; Water Resources Research 28(9)
Lund, 1988
"Metering Utility Services: Evaluation and Maintenance"; Water resourcesResearch
89
Martinez-Espineira and Nauges 2004
"Is all domestic water consumption sensitive to price control?"
Applied Economics 36, 1697-1703
Mousa Abu-Arabi and K. Venkat Reddy, 2005
R&D Challengesfor FurtherCost Reductions in Desalination
Middle East DesalinationResearch Center
Nieswiadomy, 1992
"Estimating Urban Residential Water Demand: Effects of price structure, Conservation,
and Education." Water Resources Research 28 (3)
Nieswiadomy and Molina, 1991
"A note on Price Perception in Water demand Models", Land Economics 67(3)
Nieswiadomy and Molina, 1991
"Comparing Residential Water Demand Estimates under Decreasing and Increasing
Block Rates Using Household Data."
Nauges and Thomas, 2000
Privately Operated Water Utilities, Municipal Price Negotiation, and Estimation of
Residential Water Demand: The Case of France." Land Economics 76(1)
Opaluch, 1982
"Urban Residential Demand for Water in the United States: Further discussion."
Land Economics 58(2)
Pint, 1999
"Household responses to Increased Water Rates During the California Drought."
Land Economics 75 (2)
Riazaiza, 1991
"A case of the major Cities of the Western Region of Saudi Arabia."
Land Economics 27(5)
Renwick and Archibald, 1998
"Demand Side Management Policies for Residential Water Use: Who Bears the
conservation Burden." Land Economics 74(3)
Renwick, Green, and McCorkle, 1998
"Measuring the price responsiveness of Residential Water Demand in California's Urban
Areas." A report prepared for the California department of Water resources
Renwick and Green, 1999
90
Do Residential Water Demand Side Management Policies Measure Up? An Analysis of
Eight California Water Agencies"
Journalof EnvironmentalEconomics and Management
Schefter and David, 1985
Estimating Residential Water Demand under Multi-Pat Tariffs Using Aggregate data
Land Economics 61 (3)
Qdais and Nassay, 2001
"Effect of pricing policy on water conservation: a case study"
Water policy 3(3)
AWWA, 2004
Water Desalting - Planning Guide for Water Utilities
UN Environmental Protection Program - Freshwater consumption in Europe (2006)
(http://www.unep.org/)
Zhou and Tol, 2005
Evaluating the costs of desalination and water transport,
Water Resources research (41)
Peter Rodgers,
2005 Freeman Lecture,
Woo, 1992
Drought Management, Service Interruption, and Water Pricing: Evidence from Hong Kong
Water Resources Research 28 (10)
91